A Spill Risk Assessment of the Enbridge Northern Gateway Project

A Spill Risk Assessment of the Enbridge
Northern Gateway Project
Dr. Thomas Gunton
Sean Broadbent
School of Resource and Environmental Management
Simon Fraser University
April 2013

Executive Summary
1. The objective of this report is to evaluate Enbridge’s spill risk analysis and to
determine if spills from the ENGP are likely to cause significant adverse environmental
effects as defined by the Canadian Environmental Assessment Act (CEAA).
2. The CEAA evaluation criterion requires assessment of two components to define risk:
the severity of an adverse impact and the likelihood of an adverse impact occurring.
This report focuses on evaluating the likelihood of an adverse impact (oil spills)
occurring. Another report completed by Gunton and Broadbent (2012b) addresses the
severity of the impact of an oil spill and concludes that oil spills as small as 238 m 3
(1,500 barrels) could have significant adverse environmental impacts.
3. The ENGP consists of an 1,172 kilometre (km) oil pipeline to ship oil from Alberta (AB)
to Kitimat, British Columbia (BC), a condensate pipeline to ship condensate from
Kitimat to AB, and a marine terminal for tankers. The planned capacity of the oil
pipeline is 525 thousand barrels per day (kbpd) with potential to expand to 850 kbpd
and the planned capacity of the condensate pipeline is 193 kbpd with potential to
expand to 275 kbpd. The ENGP would require an average of 220 tankers per year,
which could increase to 331 tankers per year with expansion of the pipeline to full
capacity.
4. Enbridge provides separate estimates of the likelihood of spills for each of the three
major components of the project: tanker operations, terminal operations, and the oil
and condensate pipelines. Enbridge does not combine these separate estimates to
provide an overall estimate of the probability of spills for the entire project and
therefore does not provide sufficient information to determine the likelihood of
adverse environmental effects as required by CEAA.
5. Forecasting spill risk is challenging due to the many variables impacting risk and the
uncertainties in forecasting future developments affecting risk. To improve the
accuracy of risk assessment, international best practices have been developed. This
report uses these international risk assessment best practices to evaluate Enbridge’s
methodology for estimating spill rates for the ENGP based on the following rating scale:
• Fully met: no deficiencies
• Largely met: no major deficiencies
• Partially met: one major deficiency
• Not met: two or more deficiencies.
6. The evaluation concludes that Enbridge’s spill risk analysis meets none of the seven
best practice criteria (Table ES-­‐1). In total there are 28 major deficiencies in the
Enbridge risk analysis for ENGP tanker, terminal and pipeline spills. The results show
that Enbridge understates the risk of spills from the ENGP.
i

Table ES-­‐1: Results for the Qualitative Assessment of Spill Estimates for the ENGP
Criterion Major Deficiencies Result
Transparency
Documentation fully and
effectively discloses supporting
evidence, assumptions, data
gaps and limitations, as well
as uncertainty in data and
assumptions, and their
resulting potential
implications to risk
Replicability
Documentation provides
sufficient information to allow
individuals other than those
who did the original analysis
to obtain similar results
Clarity
Risk estimates are easy to
understand and effectively
communicate the nature and
magnitude of the risk in a
manner that is complete,
informative, and useful in
decision making
Reasonableness
The analytical approach
ensures quality, integrity, and
objectivity, and meets high
scientific standards in terms of
analytical methods, data,
assumptions, logic, and
judgment
Reliability
Appropriate analytical
methods explicitly describe
and evaluate sources of
uncertainty and variability
that affect risk, and estimate
the magnitudes of
uncertainties and their effects
on estimates of risk
Validity
Independent third-­‐party
experts review and validate
findings of the risk analysis to
ensure credibility, quality, and
integrity of the analysis
Risk Acceptability
Stakeholders identify and
agree on acceptable levels of
risk in a participatory, open
decision-­‐making process and
assess risk relative to
alternative means of meeting
project objectives
1. Lack of supporting evidence for LRFP tanker incident frequency data
2. Insufficient evidence supporting conditional probabilities for tanker spills
3. Limited transparency in the application of local scaling factors to the BC
study area
4. Lack of information comparing incident frequencies at Kitimat Terminal to
marine terminals in Norway
5. Lack of transparency in documenting how mitigation measures will reduce
the likelihood of tanker and terminal spills
6. Inadequate evidence supporting the reduction of spill frequencies for
pipeline operations
Insufficient proprietary data and information required to replicate:
7. Incident frequencies for tanker accidents
8. Scaling factors for ENGP routes
9. Probabilities and spill volume estimates for tanker spills,
10. Mitigation measures for tanker and terminal operations
11. Pipeline incident frequencies
12. Failure to effectively communicate the probability of spills over the life of
the project
13. Failure to combine estimates for tanker, terminal, and pipeline spills to
present a single spill estimate for the entire project
14. Failure to determine the likelihood of all spill size categories that can have
a significant adverse environmental impact
15. Lack of clear communication of spill estimates generated in detailed
technical data reports in the main ENGP application
16. Failure to address the effective implementation of mitigation measures
that reduce risk
17. Inappropriate definition of the study area to estimate spill return periods
for tanker operations
18. Reliance on tanker incident frequency data that underreport incidents by
between 38 and 96%.
19. Potential double-­‐counting of mitigation measures for LRFP tanker incident
data
20. Inadequate consideration of the corrosiveness of diluted bitumen
associated with internal corrosion that may increase spill probability
21. Failure to consider different vessel characteristics in tanker incident
frequencies
22. Failure to provide confidence intervals that communicate uncertainty and
variability in spill estimates for tanker, terminal, and pipeline operations
23. Failure to complete an adequate sensitivity analysis that effectively
evaluates uncertainty associated with spill estimates
24. Inadequate expert review and validation of spill estimates for tanker,
terminal, and pipeline operations
25. Experts in the review and validation of spill estimates lack independence
and third-­‐party status
26. Failure to define risk acceptability in terms of the needs, issues, and
concerns of stakeholder potentially impacted by the project
27. Failure to effectively consult with stakeholders to determine a level of
acceptable risk for tanker, terminal, and pipeline spills
28. Inadequate assessment and comparison of risks associated with project
alternatives
Not
Met
Not
Met
Not
Met
Not
Met
Not
Met
Not
Met
Not
Met
ii

7. To assess the impact of deficiencies in Enbridge’s analysis of tanker spill risk estimates,
we undertake additional sensitivity analysis based on the following scenarios:
a) Combining the four sensitivity parameters that were tested separately by Enbridge
(higher scaling factors for grounding, higher ship traffic, extended route, and higher
number of tankers)
b) Using incident frequencies derived from Lloyds Register Fairplay data without the
scaling factor adjustments for BC conditions that were done by Enbridge without
sufficient documentation or justification
c) Increasing conditional probabilities by five percentage points
d) Increasing tanker traffic to transport pipeline throughput of 1,125 kbpd (850 kbpd
of oil and 275 kbpd of condensate), which is the design capacity of the ENGP
e) Increasing tanker traffic associated with replacing very large crude carriers (VLCC)
with smaller Suezmax and Aframax tankers to transport throughput of 1,125 kbpd
of oil and condensate
f) Testing the combined effect of all the above parameters.
8. Enbridge did not complete any sensitivity analysis for oil or condensate spills at
Kitimat Terminal. To address this deficiency, we undertake a sensitivity analysis for
marine terminal operations based on the following scenarios:
a) Increasing tanker traffic to transport pipeline throughput of 1,125 kbpd (850 kbpd
of oil and 275 kbpd of condensate), which is the design capacity of the ENGP
b) Replacing larger VLCCs with smaller tankers to transport throughput of 1,125 kbpd
of oil and condensate that results in an increase in the number of tanker loadings.
9. Enbridge did not complete any sensitivity analysis for oil or condensate spills from the
ENGP pipelines. To address this deficiency, we undertake a sensitivity analysis based
on the following scenarios:
a) Using actual spill rates provided by Enbridge for Enbridge’s liquids pipeline system
for the years 2002-­‐2010
b) Using spill rates from the NEB for pipebody and operational leaks exceeding 1.5 m 3
or 9 barrels (bbl).
10. Table ES-­‐2 summarizes the impacts of the sensitivity analyses. The results provide a
range of possible spill rates for the ENGP and show that the impact of the alternative
scenarios on spill rates is significant:
• The tanker spill return period is between 23 and 196 years, well below Enbridge’s
mitigated tanker spill estimate of 250 years. This lower estimate of 23 to 196 years
still underestimates spill risk because it does not correct for all the deficiencies in
Enbridge’s spill risk methodology such as underreporting of incidents.
• The terminal spill return period is between 15 and 41 years compared to Enbridge’s
estimate of 62 years.
• The pipeline spill return period of 0.1 years (between 15 and 16 spills per year)
based on the actual spill rate experienced on the Enbridge liquids pipeline system
from 2002 to 2010 is 31 times higher than the 2 year return period estimated by
Enbridge.
iii

These findings show that the deficiencies in Enbridge’s analysis result in an
underestimate of spill risk.
Table ES-­‐2: Spill Type and Sensitivity Analysis for ENGP Tanker, Terminal, and Pipeline Spills
Type (Oil or Condensate)
Enbridge Estimate
(Return period in years)
Adjusted Sensitivity
(Return period in years)
Tanker Spill 250 23 -­‐ 196
Terminal Spill 62 15 -­‐ 41
Pipeline Spill 2 0.1 -­‐ 1
Note: Return periods estimated by Enbridge and used in the adjusted sensitivity analysis represent any size tanker or terminal
spill (average spill size of 56,700 bbl and 1,575 bbl, respectively). For pipeline spills, Enbridge estimates return periods using
different datasets including reportable spill data from the US PHMSA that represents liquid releases of 5 bbl or more (Enbridge
uses average spill size of 594 bbl). The adjusted sensitivity analysis estimates return periods based on: (1) Reportable spills on the
Enbridge liquids pipeline system (average spill size of 162 bbl), and; (2) NEB pipeline spill data for leaks greater than 9 bbl.
11. An alternative methodology for estimating spill risk is the United States (US) Oil Spill
Risk Analysis (OSRA) model. The OSRA model was developed in 1975 by the United
States Department of the Interior and is the model that the US government uses to
estimate oil spill likelihood. The OSRA model has been subject to extensive evaluation
and improvements since its development and was recently updated in 2012. Despite
its widespread use, the OSRA model is not used or referenced by Enbridge. Enbridge
provides no justification for not using the OSRA model.
12. Applying the OSRA model results in probability estimates of one or more tanker spills
over 1,000 bbl of 95.3% to 99.9% (Table ES-­‐3). The OSRA estimates the probability of
one or more tanker spills over 10,000 bbl at 65.1% to 98.2%. Based on planned
shipment volumes, the OSRA model estimates oil/condensate spill return periods of 5
to 12 years for spills over 1,000 bbl, between 20 and 50 times more frequent than the
Enbridge estimate of 250 years. The OSRA model estimates oil/condensate spill return
periods of between 3 and 8 years using the higher volumes based on the ENGP
expansion capacity.
Table ES-­‐3: Spill Probabilities for ENGP Tanker Operations Based on OSRA Model
Return Period Spill Probability
(in years) over 50 years (%)
Oil spill ≥ 1,000 bbl 7 -­‐ 17 95.3 -­‐ 99.9
Oil/Condensate spill ≥ 1,000 bbl 5 -­‐ 12 98.5 -­‐ 99.9
Oil spill ≥ 10,000 bbl 13 -­‐ 48 65.1 -­‐ 98.2
OSRA Model Tanker Spill Size
Planned
Capacity
Expansion
Capacity
Oil/Condensate spill ≥ 10,000 bbl 10 -­‐ 35 76.3 -­‐ 99.6
Oil spill ≥ 1,000 bbl 4 -­‐ 11 99.3 -­‐ 99.9
Oil/Condensate spill ≥ 1,000 bbl 3 -­‐ 8 99.9
Oil spill ≥ 10,000 bbl 8 -­‐ 30 81.8 -­‐ 99.9
Oil/Condensate spill ≥ 10,000 bbl 6 -­‐ 23 89.5 -­‐ 99.9
Note: Planned capacity represents 525 kbpd of oil and 193 kbpd of condensate; Expansion capacity represents 850 kbpd of oil and
275 kbpd of condensate. The range in results of the OSRA model is based on the differences in years used for calculating spill rates.
The high end of the probability range (low end of the return period) is based on the years 1974 to 2008 and the low end of the
probability range (high end of the return period) is based on the years 1994 to 2008. The reduced risk based on the 1994 to 2008
period reflects improvements in safety mitigation and is likely more indicative of future spill rates.
iv

13. The conclusions of this report are as follows:
I. The ENGP Application Contains Major Deficiencies that Underestimate Spill Likelihood
Enbridge’s spill risk analysis contains 28 major deficiencies. As a result of these
deficiencies, Enbridge underestimates the risk of the ENGP. Some of the key
deficiencies include:
• Failure to present the probabilities of spills over the operating life of the ENGP
• Failure to evaluate spill risks outside the narrowly defined BC study area
• Reliance on LRFP data that underreport tanker incidents by between 38 and
96%.
• Failure to include the expansion capacity shipment volumes in the analysis
• Failure to provide confidence ranges of the estimates
• Failure to provide adequate sensitivity analysis
• Failure to justify the impact of proposed mitigation measures on spill likelihood
• Potential double counting of mitigation measures
• Failure to provide an overall estimate of spill likelihood for the entire ENGP
• Failure to disclose information and data supporting key assumptions that were
used to reduce spill risk estimates
• Failure to use other well accepted risk models such as the US OSRA model.
II. Deficiencies in the ENGP Application Fail to Address the CEAA Decision Criterion
Due to deficiencies in the oil spill risk assessment, Enbridge does not provide an
accurate assessment of the likelihood of significant adverse environmental impacts
as required under CEAA. Enbridge expresses the likelihood of a spill as return
periods for separate components of the project (tanker, terminal, and pipeline)
rather than the probability of a spill over the life of the project. Without estimates of
the probability of a spill over the operating life of the project, decision makers are
unable to determine whether the project meets the CEAA criterion of likelihood of
adverse environmental effects. Furthermore, due to the deficiencies in the spill risk
methodology, Enbridge fails to adequately disclose the extent of risk associated with
the ENGP and the uncertainty in estimating risk.
III. Restating Enbridge’s Own Findings as Probabilities Over Project Life Shows that Spill
Likelihood is High
When findings in the ENGP application are restated as the probability of a spill over
the operational life of the ENGP instead of return periods, the conclusion based on
Enbridge’s own analysis is that the likelihood of a spill is high (Table ES-­‐4). These
spill probabilities based on Enbridge’s analysis underestimate spill likelihood
because they are based on methodological deficiencies.
v

Table ES-­‐4: Probabilities for ENGP Tanker, Terminal, and Pipeline Spills (50 Years)
Spill Type (Oil or Condensate)
Spill Probability
over 50 years (%)
Tanker Spill 18.2
Terminal Spill 55.6
Pipeline Spill 99.9
Combined Spills 99.9
IV. Sensitivity Analysis Shows that the Likelihood of a Spill is High
The extended sensitivity analysis for the ENGP demonstrates that there is a higher
likelihood of spill occurrence than reported in the ENGP application. For tanker
spills, the extended sensitivity analysis evaluates changes to key input parameters
such as increasing shipments to the ENGP design capacity and combining parameter
changes.
a. The extended sensitivity analysis shows that based on Enbridge’s own data,
the rate of tanker spills could range between 0.3 and 2.2 spills over a 50-­‐year
period, much higher than Enbridge’s estimate of less than one spill (Figure ES-­‐
1). This range still understates spill risk because it does not correct for all the
deficiencies in Enbridge’s methodology.
Figure ES-­‐1: Number of ENGP Tanker Spills in 50 Years (Oil and Condensate)
b. The extended sensitivity analysis for terminal spills determines that the rate
of spills in a 50-­‐year period could range between 1.2 and 3.4 spills for any
size terminal spill, significantly higher than Enbridge’s estimate of less than
one spill in 50 years (Figure ES-­‐2).
vi

Figure ES-­‐2: Number of ENGP Terminal Spills in 50 Years (Oil and Condensate)
c. The extended sensitivity analysis using spill frequency data from Enbridge’s
liquid pipeline system results in an estimated 776 oil and condensate
pipeline spills in 50 years (or between 15 to 16 spills per year), which is 31
times more frequent than Enbridge’s estimate of 25 spills in 50 years (Figure
ES-­‐3).
Figure ES-­‐3: Number of ENGP Pipeline Spills in 50 Years (Oil and Condensate)
V. The OSRA Model Shows the Probability of a Major Spill is High
The OSRA model is the model the US government uses to estimate spill risk. The
OSRA model estimates 4 to 10 tanker spills over 1,000 bbl (159 m 3 ) could occur in a
50-­‐year period (Figure ES-­‐1), which is 20 to 50 times more frequent than Enbridge’s
estimate. Due to mitigation measures, the lower end of the OSRA spill range is more
likely a better indicator of future risk. In terms of spill probability, the OSRA model
estimates that an oil/condensate tanker spill ≥ 1,000 bbl has a 98.5% to 99.9%
chance of occurring over a 50-­‐year operating period, well above Enbridge’s estimate
of 18.2% (Figure ES-­‐4). The lower estimate of 98.5% is likely a better indication of
future spill risk due to mitigation measures. The probability of a tanker spill
vii

increases to 99.9% if the volume of oil and condensate handled increases to the
ENGP expansion capacity of 1,125 kbpd.
Figure ES-­‐4: Probabilities for ENGP Tanker Spills over 50 Years (Oil or Condensate)
VI. Other Considerations
Finally, two important considerations must be reconciled in the ENGP regulatory
application: risk acceptability to stakeholders, and viable alternatives to the ENGP
that reduce risk. Risk acceptability in the ENGP application relies on a technical
definition of risk. Acceptable risk however is a value judgment that should be based
on the definition of risk as defined by impacted stakeholders. Stakeholders’
definition of acceptable risk therefore needs to be assessed and used in the ENGP
decision. Second, Enbridge has not identified nor compared viable alternatives that
meet the stated purpose of the ENGP. There are viable alternatives to shipping
crude oil from the Western Canada Sedimentary Basin to market that involve no risk
from oil tanker spills and thus risk from tanker spills can be eliminated. These
alternatives should be evaluated relative to the ENGP to assess the most cost
effective transportation option.
VII. Overall Conclusion
We acknowledge that estimating risk is challenging due to the many uncertainties
involved and that different assumptions and methodologies will produce different
assessments of risk. The key in risk assessment is to test a range of assumptions
and methodological approaches to provide a clear and accurate assessment of the
range of likely outcomes so that decision makers can fully judge risk. The
conclusion of this report is that Enbridge’s oil spill risk assessment contains
methodological deficiencies and does not therefore provide an accurate assessment
of the degree of risk associated with the ENGP. The risk assessment in this report
also concludes that the ENGP has a very high likelihood of a spill that may have
significant adverse environmental effects.
viii

Table of Contents
EXECUTIVE SUMMARY ........................................................................................................................................ I
1. INTRODUCTION ........................................................................................................................................... 1
2. THE ENBRIDGE NORTHERN GATEWAY PROJECT............................................................................. 2
2.1. TANKER TRAFFIC ACCESSING KITIMAT TERMINAL...............................................................................................2
2.2. MARINE TERMINAL OPERATIONS AT KITIMAT TERMINAL ..................................................................................4
2.3. CRUDE OIL AND CONDENSATE PIPELINE.................................................................................................................5
3. SUMMARY OF METHODOLOGY AND SPILL ESTIMATES FOR THE ENGP ................................... 5
3.1. SPILLS FROM TANKER TRAFFIC ACCESSING KITIMAT TERMINAL.......................................................................6
3.2. SPILLS FROM OPERATIONS AT KITIMAT TERMINAL ..............................................................................................9
3.3. SPILLS FROM PIPELINE OPERATIONS .................................................................................................................... 12
3.4. SUMMARY OF TANKER, TERMINAL, AND PIPELINE SPILL ESTIMATES FOR THE ENGP............................... 15
4. EVALUATION OF ENGP SPILL ESTIMATES.........................................................................................16
4.1. SUMMARY OF QUALITATIVE ASSESSMENT OF SPILL ESTIMATES FOR THE ENGP........................................ 43
5. EXTENDED SENSITIVITY ANALYSIS FOR ENGP SPILL ESTIMATES............................................45
5.1. SENSITIVITY ANALYSIS FOR SPILL RETURN PERIODS FOR TANKER OPERATIONS........................................ 45
5.2. SENSITIVITY ANALYSIS FOR SPILL RETURN PERIODS FOR OPERATIONS AT KITIMAT TERMINAL............. 49
5.3. SENSITIVITY ANALYSIS FOR SPILL RETURN PERIODS FOR PIPELINE OPERATIONS...................................... 50
5.4. SUMMARY OF EXTENDED SENSITIVITY ANALYSIS FOR ENGP TANKER AND TERMINAL OPERATIONS.... 53
6. APPLICATION OF OIL SPILL RISK ANALYSIS MODEL TO THE ENGP .........................................54
6.1. OVERVIEW OF OIL SPILL RISK ANALYSIS MODEL ............................................................................................... 54
6.2. SPILL PROBABILITY ESTIMATES FOR ENGP TANKER OPERATIONS ............................................................... 56
6.3. SPILL PROBABILITY ESTIMATES FOR ENGP PIPELINE ...................................................................................... 61
6.4. SUMMARY OF ENGP SPILL PROBABILITIES ESTIMATED WITH OSRA MODEL............................................. 63
6.5. EVALUATION OF OSRA MODEL WITH BEST PRACTICE CRITERIA FOR RISK ASSESSMENT......................... 63
6.6. SUMMARY OF FINDINGS FOR QUALITATIVE ASSESSMENT OF OSRA MODEL ................................................ 68
7. COMPARISON OF SPILL ESTIMATES FOR THE ENGP......................................................................69
8. CONCLUSION ...............................................................................................................................................73
REFERENCES........................................................................................................................................................77
APPENDIX A -­‐ DETAILED SUMMARY OF ENGP SPILL RISK ANALYSIS ..............................................84
ix

List of Equations, Figures and Tables
EQUATION 1: OSRA MODEL EQUATION (POISSON ASSUMPTION)..................................................................................................................................................................56
FIGURE 1: MAP OF ENGP TANKER OPERATIONS...................................................................................................................................................................................................4
FIGURE 2: HISTORICAL SPILL PERFORMANCE OF THE ENBRIDGE LIQUIDS PIPELINE SYSTEM (1998 -­‐ 2010) .....................................................................................23
FIGURE 3: SUMMARY OF SENSITIVITY ANALYSIS FOR ENGP TANKER, TERMINAL, AND PIPELINE SPILLS..............................................................................................54
FIGURE 4: NUMBER OF ENGP TANKER SPILLS IN 50 YEARS (OIL AND CONDENSATE)..............................................................................................................................74
FIGURE 5: NUMBER OF ENGP TERMINAL SPILLS IN 50 YEARS (OIL AND CONDENSATE)..........................................................................................................................75
FIGURE 6: NUMBER OF ENGP PIPELINE SPILLS IN 50 YEARS (OIL AND CONDENSATE) ............................................................................................................................75
FIGURE 7: PROBABILITIES FOR ENGP TANKER SPILLS OVER 50 YEARS (OIL OR CONDENSATE)..............................................................................................................76
TABLE 1: CHARACTERISTICS OF OIL AND CONDENSATE TANKERS ACCESSING KITIMAT TERMINAL...........................................................................................................2
TABLE 2: LRFP BASE INCIDENT FREQUENCIES AND LOCAL SCALING FACTORS FOR INCIDENTS IN BC STUDY AREA .............................................................................7
TABLE 3: CONDITIONAL SPILL PROBABILITIES FOR TANKER INCIDENTS (METHOD 1 AND METHOD 2) ...................................................................................................7
TABLE 4: SUMMARY OF SENSITIVITY ANALYSIS ON SPILL RETURN PERIODS (YEARS)...................................................................................................................................8
TABLE 5: ESTIMATED REDUCTION IN THE FREQUENCY OF INCIDENTS FROM TUG USE .................................................................................................................................9
TABLE 6: COMPARISON OF UNMITIGATED AND MITIGATED OIL/CONDENSATE SPILL RETURN PERIODS..................................................................................................9
TABLE 7: COMPARISON OF UNMITIGATED AND MITIGATED RETURN PERIODS FOR VARIOUS OIL SPILL SIZES ........................................................................................9
TABLE 8: INCIDENTS TYPES EXAMINED FOR MARINE TERMINAL SPILLS ......................................................................................................................................................10
TABLE 9: SPILL PROBABILITY AND SPILL SIZE DISTRIBUTION FOR CARGO TRANSFER OPERATIONS.......................................................................................................10
TABLE 10: COMPARISON OF OVERALL UNMITIGATED AND MITIGATED OIL/CONDENSATE SPILL RETURN PERIODS FOR CARGO TRANSFER OPERATIONS........11
TABLE 11: RUPTURE FAILURE FREQUENCIES FOR MARINE TERMINAL COMPONENTS...............................................................................................................................11
TABLE 12: RETURN PERIODS FOR SPILLS FROM THE NORTHERN GATEWAY PIPELINE ..............................................................................................................................13
TABLE 13: RETURN PERIODS FOR LEAKS AND FULL-­‐BORE RUPTURES ..........................................................................................................................................................14
TABLE 14: FREQUENCIES FOR PIPING AND PUMP SPILLS AT SMOKY RIVER STATION.................................................................................................................................14
TABLE 15: NEB PIPELINE FAILURE FREQUENCIES ADJUSTED DOWNWARD TO REPRESENT THE ENGP ...............................................................................................15
TABLE 16: SUMMARY OF SPILL RETURN PERIODS FOR THE ENGP.................................................................................................................................................................16
TABLE 17: EVALUATIVE CRITERIA FOR ASSESSING THE QUALITY OF METHODOLOGICAL APPROACHES TO ESTIMATING RISK..........................................................17
TABLE 18: RECENT SPILL RECORD FOR THE KEYSTONE PIPELINE (2010 -­‐ 2011)....................................................................................................................................24
TABLE 19: SPILL PROBABILITIES OVER THE LIFE OF THE ENGP BASED ON REGULATORY APPLICATION..............................................................................................27
TABLE 20: DIFFERENCES IN NAUTICAL MILES USED TO ESTIMATE TANKER SPILLS AND ACTUAL NAUTICAL MILES TO KEY EXPORT MARKETS..........................32
TABLE 21: ESCORT TUG USAGE AT INTERNATIONAL TANKER PORTS............................................................................................................................................................34
TABLE 22: RESULTS FOR THE QUALITATIVE ASSESSMENT OF SPILL ESTIMATES FOR THE ENGP............................................................................................................44
TABLE 23: SENSITIVITY ANALYSIS COMBINING SENSITIVITY PARAMETERS IDENTIFIED BY DNV............................................................................................................46
TABLE 24: SENSITIVITY ANALYSIS BASED ON UNSCALED INCIDENT FREQUENCIES.....................................................................................................................................46
TABLE 25: FIVE-­‐PERCENTAGE POINT INCREASE IN CONDITIONAL PROBABILITIES FOR SENSITIVITY ANALYSIS ..................................................................................47
TABLE 26: SENSITIVITY ANALYSIS FOR INCREASED CONDITIONAL PROBABILITIES BY FIVE PERCENTAGE POINTS..............................................................................47
TABLE 27: POTENTIAL EXPANSION SCENARIO OF OIL AND CONDENSATE PIPELINES.................................................................................................................................47
TABLE 28: SENSITIVITY ANALYSIS FOR TANKERS TRANSPORTING 1,125 KBPD..........................................................................................................................................48
TABLE 29: SENSITIVITY ANALYSIS FOR SMALLER TANKERS TRANSPORTING THROUGHPUT OF 1,125 KBPD........................................................................................48
TABLE 30: SENSITIVITY ANALYSIS AGGREGATING PARAMETERS FROM THE EXTENDED SENSITIVITY ANALYSIS..................................................................................49
TABLE 31: SENSITIVITY ANALYSIS FOR INCREASED MARINE TERMINAL THROUGHPUT OF 1,125 KBPD................................................................................................50
TABLE 32: SENSITIVITY ANALYSIS FOR SMALLER TANKERS TRANSPORTING THROUGHPUT OF 1,125 KBPD AT THE MARINE TERMINAL......................................50
TABLE 33: HISTORICAL SPILL DATA FOR THE ENBRIDGE PIPELINE SYSTEM (2002 -­‐ 2010) ..................................................................................................................51
TABLE 34: SENSITIVITY ANALYSIS FOR ENGP PIPELINES USING SPILL DATA FOR THE ENBRIDGE LIQUIDS PIPELINE SYSTEM ........................................................52
TABLE 35: HISTORICAL NEB PIPEBODY AND OPERATIONAL LEAK DATA FROM 2000 TO 2009 ............................................................................................................52
TABLE 36: SENSITIVITY ANALYSIS FOR ENGP PIPELINES USING NEB DATA FOR LEAKS >1.5M 3...........................................................................................................53
TABLE 37: INTERNATIONAL AND ALASKA OIL TANKER SPILL RATES ............................................................................................................................................................57
TABLE 38: ENGP TANKER SPILL PROBABILITIES BASED ON OSRA MODEL (30 YEARS) ..........................................................................................................................58
TABLE 39: ENGP TANKER SPILL PROBABILITIES BASED ON OSRA MODEL (50 YEARS) ..........................................................................................................................59
TABLE 40: SENSITIVITY ANALYSIS FOR ENGP TANKER SPILL PROBABILITIES AT HIGHER CAPACITIES (30 YEARS)..........................................................................59
TABLE 41: SENSITIVITY ANALYSIS FOR ENGP TANKER SPILL PROBABILITIES AT HIGHER CAPACITIES (50 YEARS)...........................................................................60
TABLE 42: SPILL RATES FOR PIPELINE SPILLS IN ALASKA BASED ON ANDERSON AND LABELLE (2000)..............................................................................................61
TABLE 43: PIPELINE SPILL PROBABILITIES BASED ON OSRA MODEL (30 AND 50 YEARS)......................................................................................................................62
TABLE 44: SENSITIVITY ANALYSIS FOR PIPELINE SPILL PROBABILITIES AT HIGHER CAPACITIES (30 AND 50 YEARS) ......................................................................62
TABLE 45: SUMMARY OF TANKER AND PIPELINE OIL/CONDENSATE SPILL PROBABILITIES FOR THE ENGP........................................................................................63
TABLE 46: RESULTS FOR THE EVALUATION OF THE OSRA MODEL WITH BEST PRACTICES FOR RISK ASSESSMENT ...........................................................................68
TABLE 47: COMPARISON OF METHODOLOGIES ESTIMATING SPILL LIKELIHOOD IN ENGP RISK ANALYSES AND THE OSRA MODEL ..............................................70
TABLE 48: COMPARISON OF RETURN PERIODS AND SPILL PROBABILITIES FOR THE ENGP......................................................................................................................71
TABLE 49: SUMMARY AND COMPARISON OF RESULTS FOR THE QUALITATIVE ASSESSMENT ....................................................................................................................72
TABLE 50: PROBABILITIES FOR ENGP TANKER, TERMINAL, AND PIPELINE SPILLS (50 YEARS).............................................................................................................74
x

List of Acronyms
AB Alberta
Bbbl Billion Barrels
bbl Barrels
BC British Columbia
BOEM Bureau of Ocean Energy Management
BSEE Bureau of Safety and Environmental Enforcement
CEAA Canadian Environmental Assessment Act
Dilbit Diluted Bitumen
DNV Det Norske Veritas
ENGP Enbridge Northern Gateway Project
kbpd Thousand Barrels per Day
km Kilometres
LRFP Lloyds Register Fairplay
NEB National Energy Board
nm Nautical Miles
OSRA Oil Spill Risk Analysis
PHMSA Pipeline and Hazardous Materials Safety Administration
QRA Quantitative Risk Analysis
US United States
VLCC Very Large Crude Carrier
Acknowledgements
Funding for this project was provided by Coastal First Nations. We would like to thank the
external reviewers who evaluated earlier drafts of this report and provided helpful
comments and suggestions.
xi

1. Introduction
This report provides a review of spill estimates for tanker, terminal, and pipeline
operations associated with the Enbridge Northern Gateway Project (ENGP). The objective
of the report is to assess whether the ENGP is likely to cause significant adverse
environmental effects as defined by the Canadian Environmental Assessment Act (CEAA).
According to the CEAA the decision maker must decide if the project:
(a) is likely to cause significant adverse environmental effects referred to in
subsection 5(1); and
(b) is likely to cause significant adverse environmental effects referred to in
subsection 5(2) [emphasis added] (CEAA Sec. 52).
Thus a key criterion in the decision of whether or not to approve a project is the likelihood
that the project will cause significant adverse environmental effects. In the case of the
ENGP, a fundamental component of determining the likelihood that a project will cause
significant adverse environmental effects is an assessment of the probability that a spill will
occur.
This report assesses the methodologies used to estimate spill probabilities in the ENGP
regulatory application. The study uses the following approach to achieve our research
objective:
I. Summarize spill estimates submitted by Enbridge for the ENGP
II. Evaluate the marine and terrestrial spill estimates submitted by Enbridge for the
ENGP against international best practices criteria
III. Undertake sensitivity analysis to test the impact of deficiencies on the spill rate
estimates
IV. Apply the United States (US) Oil Spill Risk Analysis (OSRA) model to estimate
marine and terrestrial spill probabilities for the ENGP
V. Submit a draft report summarizing the findings for review by international risk
assessment experts and revise report based on the review.
Following the introduction, the second section of our report summarizes the Northern
Gateway Project. The third section summarizes Enbridge’s estimate of the likelihood of
tanker, terminal, and pipeline spills. In the fourth section, we evaluate Enbridge’s spill
estimate methodology and complete a sensitivity analysis to illustrate the range of spill
rates under different scenarios. In the sixth section, we estimate alternative spill
probabilities for tanker and pipeline operations using the US OSRA model. The seventh
section compares the methodological approach and results of the OSRA model to those
submitted as part of the ENGP regulatory application and section eight provides
conclusions based on our findings.
1

It is important to note that the CEAA evaluation criterion requires assessment of two
components to define risk: the magnitude of an adverse impact and the likelihood of an
adverse impact occurring. This report focuses on evaluating only the one component of
risk: the likelihood of an adverse impact (oil spills) occurring. Another report completed
by Gunton and Broadbent (2012b) addresses the magnitude of the impact of an oil spill and
concludes that oil spills as small as 238 m 3 could have significant adverse environmental
impacts. The Gunton and Broadbent (2012b) report also concludes that Enbridge’s
analysis of the magnitude of impacts of oil spills is deficient.
2. The Enbridge Northern Gateway Project
The following section provides a brief overview of tanker traffic, marine terminal, and
pipeline operations associated with the ENGP.
2.1. Tanker Traffic Accessing Kitimat Terminal
Tankers accessing Kitimat would include oil tankers exporting crude oil to international
markets and condensate tankers importing condensate from abroad. Table 1 provides a
breakdown of characteristics for very large crude carriers (VLCC), Suezmax, and
Aframax tankers that would transport oil and condensate to and from the marine
terminal. The ENGP application forecasts an additional 190 to 250 tankers a year, or an
average of 220 vessels, to existing commercial marine traffic accessing Kitimat
(Enbridge 2010b Vol. 8B p. 2-­‐9). Tanker traffic of 220 vessels equates to 440 tanker
sailings or approximately 8 tanker transits every week. Of the 220 tankers per year
forecast to call at the proposed Kitimat Terminal, approximately 149 tankers would
export crude oil and 71 tankers would import condensate (Brandsæter and Hoffman
2010). Tankers importing condensate would be laden inbound, while tankers exporting
crude oil would be laden for outbound transits. The proposed project has capacity to
increase shipments from 525 thousand barrels per day (kbpd) of oil (83,400 m 3 per
day) to 850 kbpd and increase the condensate pipeline from 193 kbpd (30,700 m 3 per
day) to 275 kbpd with additional pumping capacity (Enbridge 2010b Information
Request No. 3). The increased shipments would increase tanker traffic from 220 to 331
per year 1 . Enbridge does not include an assessment of the impact of this higher tanker
traffic on spill risk in its analysis.
Table 1: Characteristics of Oil and Condensate Tankers Accessing Kitimat Terminal
Characteristic
VLCC
Tanker Class
Suezmax Aframax
Maximum Deadweight Tonnage 320,000 160,000 81,000
Overall Length (m) 343.7 274.0 220.8
Average Cargo Capacity (m3) 330,000 160,000 110,000
Average Number of Vessels per Year
Source: Enbridge (2010b).
50 120 50
1 Estimated based on the increase in tanker traffic according to tanker data and information provided by
Brandsæter and Hoffman (2010) in the marine shipping QRA and Dudding (2013)
2

Enbridge expects that tankers calling at Kitimat Terminal would be chartered, or owned
by independent tanker operators, and not owned or operated by Enbridge (Enbridge
2010b Vol. 8A p. 4-­‐3). Owners of chartered tankers bear the responsibility for tanker
safety. According to Enbridge (2010b Vol. 8B p. 2-­‐5), operational protocols for tankers
include local pilots, escort and tethered tugs, tankers equipped with navigation systems,
and the use of radar systems. Enbridge states that it would require all chartered
tankers for the ENGP to be double-­‐hulled (Enbridge 2010b Vol. 8A p. 4-­‐84).
Tanker traffic accessing the marine terminal at Kitimat would use three potential routes
within an area referred to as the British Columbia (BC) study area. The BC study area
includes an open water area defined as marine waters from the Alaskan border to the
northern end of Vancouver Island, and from the 12 nautical mile (nm) limit of the
Territorial Sea of Canada landward to the northern fjords, as well as a confined channel
assessment area defined as the Kitimat Arm, Douglas Channel and channels leading
from Douglas Channel to open waters of Queen Charlotte Sound and Hecate Strait. The
three tanker routes include a Northern Approach and two Southern Approaches (Figure
1). The Northern Approach is a 251 nm route that approaches north of Haida Gwaii,
continues through Hecate Strait and Wright Sound and enters Douglas Channel on its
way to Kitimat Terminal. The 190 nm Southern Approach passes directly through
Caamano Sound passage through Queen Charlotte Sound, through Hecate Strait and
Wright Sound and into Douglas Channel. Finally, the Southern Approach via Principe
Channel is a 259 nm route that enters Principe Channel north of Banks Island, through
Nepean Sound and Otter Channel, and follows a similar route as the other Southern
route into Kitimat Terminal. The three tanker routes will traverse the Pacific North
Coast Integrated Management Area and traditional territories of Coastal First Nations.
In March 2013, Transport Canada announced amendments to the Canada Shipping Act
in an attempt to strengthen Canada’s tanker safety system. The eight new measures to
the tanker safety program include:
• Conduct annual tanker inspections to ensure that foreign vessels comply with
rules and regulations
• Expand the aerial surveillance and monitoring of ships
• Establish a Canadian Coast Guard Incident Command System to respond to
incidents
• Review current pilotage and tug escort requirements
• Designate more public ports, particularly Kitimat
• Conduct scientific research on petroleum products
• Ensure the installation and maintenance of aids to navigation such as buoys and
lights
• Develop options to enhance current navigation system (TC 2013).
Only one of the eight measures, namely annual tanker inspections, represents an
actionable preventative measure to reduce the likelihood of a spill and inspections are
presently carried out on a regular basis. The existing marine safety framework in
Canada consists of Transport Canada regulations and standards under the Canada
3

Shipping Act as well as international regulations from the International Maritime
Organization. Transport Canada’s tanker policy requires all foreign tankers to be
inspected on their first visit to Canada and at least once per year thereafter (TC 2010).
The Canadian tanker inspection policy was implemented following the 1990 Brander-­‐
Smith report that was commissioned after the Exxon Valdez oil spill. Canada is also a
member of international Port State Control agreements enabled by the International
Maritime Organization to ensure that foreign vessels comply with international
maritime conventions. Canada is among the 27 signatories of the Paris Memorandum
of Understanding since 1994 and is among the 18 signatories of the Tokyo
Memorandum of Understanding since 1993, both of which entitle member states to
inspect foreign ships entering their waters. Thus tankers entering Canada’s territorial
waters not only undergo regular inspection in Canada but are also inspected by many
other nations that are members of the international tanker inspection regime.
Figure 1: Map of ENGP Tanker Operations
Source: Living Oceans (2011)
2.2. Marine Terminal Operations at Kitimat Terminal
Kitimat Terminal is a proposed 220-­‐hectare marine terminal complex on Kitimat Arm
that would consist of a tank terminal and marine terminal. The marine terminal would
have two principal functions; (1) receive crude oil from the oil pipeline, transfer it to
tanks at the marine terminal and load oil onto tankers, and; (2) unload and transfer
condensate from tankers to tanks and into the condensate pipeline (Enbridge 2010b
Vol. 3).
4

Major facilities at Kitimat Terminal would include oil and condensate transfer systems,
terminal buildings, and tanker berths. The oil transfer system would include an oil
receiving station that will direct oil into storage tanks, 11 oil tanks with a nominal
capacity of 496,000 barrels (bbl), or 78,800 m 3 , and an oil loading system that consists
of loading arms for loading oil into tankers (Enbridge 2010b Vol. 3). The condensate
transfer system would receive condensate from tankers and pump it to three tanks
with similar capacity characteristics as oil tanks, after which the initiating condensate
pump station will direct condensate into the pipeline for transport to Bruderheim
Station. Approximately 45 buildings are planned for Kitimat Terminal, including
various maintenance and storage buildings, electrical buildings, and pumphouses
(Enbridge 2010b Vol. 3). The terminal would also have two tanker berths to transfer
oil and condensate into and out of VLCC, Suezmax, and Aframax tankers.
2.3. Crude Oil and Condensate Pipeline
Separate crude oil and condensate pipelines would transport hydrocarbons between
Kitimat, BC and Bruderheim, Alberta (AB). The crude oil pipeline would deliver diluted
bitumen (dilbit) and synthetic oil in the export pipeline from AB to Kitimat Terminal
where it will be loaded onto tankers and shipped to market. The condensate pipeline
would import condensate from tankers arriving at Kitimat Terminal and deliver it to
AB where condensate will be used for blending with crude oil (Enbridge Vol. 2 p. 1-­‐12).
Construction and operation of the oil and condensate pipelines would consist of 1,172
kilometres (km) of pipe, seven pump stations for the oil pipeline, nine pump stations
for the condensate pipeline, and Bruderheim Station that would house an initiating oil
pump station and condensate receiving facilities (Enbridge 2010b Vol. 3). The crude
oil pipeline is designed for an initial average annual throughput of 525 kbpd of oil
(83,400 m 3 per day) with potential to expand to 850 kbpd with additional pumping
capacity and the condensate pipeline is designed for an initial average annual
throughput of 193 kbpd (30,700 m 3 per day) with potential to expand to 275 kbpd
(Enbridge 2010b Information Request No. 3). The pipelines will traverse several
different physiographic regions including plains, plateaus, and mountains and will
cross 773 watercourses, 669 of which are fish-­‐bearing (Enbridge 2010b Vol. 3). The
traditional territories of approximately 50 First Nations are along the pipeline route
(Hoekstra 2012).
3. Summary of Methodology and Spill Estimates for the ENGP
The following section provides a brief summary of the methodologies used by Enbridge
and its consultants to estimate spill likelihood for ENGP tanker, terminal, and pipeline spills
(Appendix A contains a more detailed summary). Methodologies and spill estimates in the
ENGP regulatory application provide context for the evaluation in section four.
5

Separate quantitative analyses estimating return periods 2 for tanker, terminal, and pipeline
spills were prepared for the ENGP regulatory application. Enbridge contracted Det Norske
Veritas (DNV) to prepare the Technical Data Report entitled Marine Shipping Quantitative
Risk Analysis (marine shipping QRA) that examines tanker and terminal spills in the BC
study area and spills at the marine terminal in Kitimat (Brandsæter and Hoffman 2010).
Furthermore, Enbridge prepared the risk analysis for pipeline spills between Kitimat, BC
and Bruderheim, AB (Enbridge 2010b Vol. 7B) included in Volume 7B: Risk Assessment and
Management of Spills -­‐ Pipelines. An information request from the Joint Review Panel
resulted in additional reports submitted in May 2012 evaluating risks associated with
Kitimat Terminal (Bercha Group 2012a), public safety in populated areas along the pipeline
route (Bercha Group 2012b), pipeline pump stations (Bercha Group 2012c), and pipeline
leaks and ruptures (WorleyParsons 2012). The summary of spill estimates for the ENGP in
this section largely relies on the following documents submitted as part of the ENGP
regulatory application:
• Volume 7B: Risk Assessment and Management of Spills -­‐ Pipelines
• Volume 7C: Risk Assessment and Management of Spills -­‐ Kitimat Terminal
• Volume 8C: Risk Assessment and Management of Spills -­‐ Marine Transportation
• TERMPOL Study No. 3.15: General Risk Analysis and Intended Methods of Reducing
Risk
• Marine Shipping Quantitative Risk Analysis by Det Norske Veritas
• Northern Gateway Pipeline Kitimat Terminal Quantitative Risk Analysis by Bercha
Group
• Northern Gateway Pipeline Public Safety Quantitative Risk Analysis by Bercha Group
• Northern Gateway Pipeline Pump Station Quantitative Risk Analysis by Bercha Group
• Semi-­‐Quantitative Risk Assessment by WorleyParsons.
3.1. Spills from Tanker Traffic Accessing Kitimat Terminal
DNV determines spill return periods for tanker traffic accessing Kitimat Terminal in the
Marine Shipping Quantitative Risk Analysis. DNV uses a per-­‐voyage methodology to
forecast spills in the marine shipping QRA based on nautical miles (nm) travelled by
tankers within the study area. As part of the per-­‐voyage methodology, DNV determines
base incident frequencies from historical tanker incident data from Lloyds Register
Fairplay (LRFP) for the period 1990-­‐2006 (Brandsæter and Hoffman 2010 p. 5-­‐49).
The LRFP data likely reflects tankers operating in jurisdictions where mitigation
measures such as escort tugs and marine safety measures such as those announced by
the government of Canada are already in use. However, a lack of transparency in the
marine QRA prevents verification of the LRFP data.
DNV then calculates incident data for the BC study area by multiplying the international
frequency data from LRFP by scaling factors that compare risks in the study area to the
international areas on which the incident data are based. The comparison uses 1.0 as a
baseline and thus a scaling factor equal to 1.0 suggests that the local conditions
2 DNV defines a return period as the number of years between spill events (Brandsæter and Hoffman 2010 p.
7-­‐94).
6

affecting risk are equal to the world average (Brandsæter and Hoffman 2010 p. 5-­‐52).
DNV uses a range of scaling factors including the number of course changes, use of
pilots with local knowledge, and the distance from ship to shore, among others. Table 2
presents base LRFP incident frequencies and the range of local scaling factors for each
incident type.
Table 2: LRFP Base Incident Frequencies and Local Scaling Factors for Incidents in BC Study Area
Tanker Incident Type
Base LRFP Incident
Frequency (per nm)
Range of Total Scaling
Factors
Powered Grounding 5.98E-­‐07 0.001 -­‐ 1.94
Drift Grounding 9.96E-­‐08 0.01 -­‐ 1.56
Collision 4.54E-­‐07 0.01 -­‐ 0.54
Foundering 5.04E-­‐10 0.01 -­‐ 1.50
Fire/Explosion
Source: Brandsæter and Hoffman (2010).
3.26E-­‐08 n/a
DNV then determines the probability of a spill resulting from a tanker incident by using
two different methods to estimate conditional spill probabilities. The first method
(Method 1) determines conditional spill probabilities based on LFRP data. According to
Brandsæter and Hoffman (2010), LRFP categorizes damages for the four incident types
into minor damage, major damage, and total loss and LRFP provides a frequency
distribution for each damage category. The second method (Method 2) estimates spill
quantities for bottom and side damages based on information from the International
Marine Organization International Convention for the Prevention of Pollution from
Ships (MARPOL) and estimates spill size distribution with Naval Architecture Package
software. Table 3 compares conditional spill probabilities from Methods 1 and 2 and
shows the mean oil outflow for grounding and collision incidents estimated with
Method 2.
Table 3: Conditional Spill Probabilities for Tanker Incidents (Method 1 and Method 2)
Incident Type
Method 1
Conditional Spill Probability (%)
Method 2
Laden Ballast Laden Ballast Mean Oil Outflow* (m3) Grounding 32.7 6.4 17.5 -­‐ 18.7 < 0.1 -­‐ 0.2 736 -­‐ 1,725
Collision 19.1 2.6 21.0 -­‐ 27.7 5.7 -­‐ 14.2 638 -­‐ 1,399
Foundering 100 100 n/a n/a
Fire/Explosion 27.0 7.6 n/a n/a
Source: Brandsæter and Hoffman (2010)
* Mean oil outflow represents outflow from laden vessels (not vessels in ballast)
In addition to determining spill probabilities for various tanker sizes and oil outflows,
Method 2 also estimates probabilities for grounding and collision spills exceeding a
certain volume. According to the Monte Carlo simulation, the probability of a spill
greater than 10,000 m 3 for a laden VLCC grounding is 5.5% and 25% for a collision
(Brandsæter and Hoffman 2010 p. 6-­‐81 -­‐ 6-­‐85). The probability of a grounding spill
exceeding 25,000 m 3 and 40,000 m 3 is 1% and 0.2%, respectively, while the probability
7

of a collision spill exceeding 20,000 m 3 and 50,000 m 3 is 10% and 0.3%, respectively
(Brandsæter and Hoffman 2010 p. 6-­‐81 -­‐ 6-­‐85).
DNV estimates unmitigated spill return periods for each type of tanker incident based
on scaled incident frequencies, conditional probabilities, the length of each segment,
and the number of times the route is travelled per year. DNV presents unmitigated risk
as a return period, which is the number of years between spill events, and is an
alternative to stating the annual probability of a spill. According to DNV, an
unmitigated tanker incident will result in an oil/condensate spill once every 78 years
(see Table 6). DNV also estimates unmitigated return periods for various oil spill sizes
and, based on Method 2 of the consequence assessment, DNV estimates that a spill
exceeding 5,000 m 3 will occur every 200 years while a spill exceeding 40,000 m 3 , will
occur every 12,000 years (see Table 7).
DNV conducts a sensitivity analysis in its marine shipping QRA examining impacts of
several parameters on spill probabilities. The sensitivity analysis tests the impact of
changes in parameters on all three potential routes separately under the assumption
that all 220 tankers forecast to call at Kitimat Terminal use the route being tested
exclusively. Therefore, the spill probabilities do not account for using a combination of
routes. Results from the sensitivity analysis suggest that return periods change
modestly compared to baseline return periods (Table 4).
Table 4: Summary of Sensitivity Analysis on Spill Return Periods (years)
Sensitivity Parameter North Route
South Route
via
Caamano
Sound
South Route
via
Browning
Entrance
Baseline 69 83 84
Sensitivity 1: Increase in Scaling Factors for Grounding 59 71 71
Sensitivity 2: Increase in Traffic Density for Collisions 67 81 82
Sensitivity 3a: Decrease to 190 Tankers per year 80 96 97
Sensitivity 3b: Increase to 250 Tankers per year 61 73 74
Sensitivity 4: 200 nm Extension to Segments 5 and 8* 67 81 82
* DNV does not provide overall return periods for the trip extension sensitivity analysis and thus return periods were calculated
based on Brandsæter and Hoffman (2010).
The next step in the DNV analysis estimates the impact of various mitigation measures
on spill return periods. DNV examines several risk reduction measures qualitatively
including enhanced navigational aids, vessel traffic management system, environmental
limits for safe operation, and the establishment of places of refuge along tanker routes.
The marine shipping QRA only quantitatively evaluates a single mitigation measure: the
tug escort plan. DNV obtains the risk reducing effect of escort tugs from a confidential
study it completed in 2002 and applies the downward adjustments directly to powered
grounding, drift grounding, and collision incidents for the ENGP (Table 5).
8

Table 5: Estimated Reduction in the Frequency of Incidents from Tug Use
Incident type Condition Reduction Effect
Powered
Grounding
Laden with close and tethered tug
Laden with close escort
Ballast with close escort
80%
Drift
Grounding
Laden with close and tethered tug
Laden with close escort
Ballast with close escort
90%
80%
Collision Laden or ballast with close and/or tethered escort 5%
Source: Brandsæter and Hoffman (2010 p. 8-­‐120).
Table 6 compares mitigated return periods to unmitigated return periods for oil or
condensate spills. For all tanker spills, mitigation measures increase the spill return
period from 78 years to 250 years. Mitigation measures for powered and drift
grounding increase return periods by three-­‐ and four-­‐fold, respectively.
Table 6: Comparison of Unmitigated and Mitigated Oil/Condensate Spill Return Periods
Incident Type
Unmitigated
Return Period
(in years)
Mitigated
Return Period
(in years)
Powered Grounding* 107 480
Drift Grounding* 535 2,758
Collision* 1,084 1,138
Foundering* 25,821 25,821
Fire/ Explosion* 1,838 1,838
Total 78 250
Source: Brandsæter and Hoffman (2010).
* Calculated based on Brandsæter and Hoffman (2010).
Note: DNV does not identify mitigation measures for foundering and fire/explosion spills.
DNV also estimates the impact of mitigation measures on return periods by size of the
spill. As shown in Table 7, the use of tugs increases return periods from 200 to 550
years for an oil spill greater than 5,000 m 3 .
Table 7: Comparison of Unmitigated and Mitigated Return Periods for Various Oil Spill Sizes
Spill Volume (m 3)
Unmitigated
Return Period
(in years)
Mitigated
Return Period
(in years)
> 5,000 200 550
> 20,000 1,750 2,800
> 40,000 12,000 15,000
Source: Brandsæter and Hoffman (2010 p. 7-­‐110; 137).
3.2. Spills from Operations at Kitimat Terminal
DNV determines spill return periods for particular types of incidents at Kitimat
Terminal in the Marine Shipping Quantitative Risk Analysis. Furthermore, the Bercha
Group estimates risks from spills to the public and workers at the terminal in the
Northern Gateway Kitimat Terminal Quantitative Risk Analysis.
9

In its risk analysis, DNV determines incident frequencies for berthing/deberthing and
cargo transfer operations at the marine terminal based on LRFP data and DNV’s
research of operating terminals in Norway. DNV examines four types of potential
incidents at the marine terminal and determines that spills associated with cargo
transfer operations pose the greatest spill risk (Table 8). Although Brandsæter and
Hoffman (2010) acknowledge that there is some risk of either a tanker striking the pier
during berthing or a passing vessel impacting a tanker, these incidents are not included
in the unmitigated return periods for terminal spills.
Table 8: Incidents Types Examined for Marine Terminal Spills
Terminal Incident Type Spill Likelihood Description
Tanker Impacted by Tug n/a
Harbour tugs will not achieve the speeds
necessary to penetrate tanker hulls
Tanker Strikes Pier
During Berthing
Low
Energy produced from pier impact is not sufficient
to penetrate tanker hull
Tanker Impacted by
Passing Vessel
Low
Risk is low due to low volume of vessel traffic
passing Kitimat Terminal
Cargo Transfer
Operations
Various
Spill risk depends on the type of transfer
operation failure
Source: Brandsæter and Hoffman (2010)
Probabilities and distribution of cargo released per loading/discharging operation are
summarized in Table 9. DNV develops a spill size distribution for each event type from a
review of historical databases from Norway and the UK and categorizes spill sizes
according to the International Tanker Owners Pollution Federation Limited 3
(Brandsæter and Hoffman 2010 p. 6-­‐91).
Table 9: Spill Probability and Spill Size Distribution for Cargo Transfer Operations
Event Type
Probability of
Spill
(per operation)
Distribution of Spill Size
Small
(< 10 m3) Medium
(10 – 1,000 m3) Release from Loading Arm 5.1E-­‐05 90% 10%
Equipment Failure 5.1E-­‐06 0% 100%
Failure in Vessels Piping System or Pumps 7.2E-­‐06 90% 10%
Human Failure 7.2E-­‐06 90% 10%
Mooring Failure 3.8E-­‐06 0% 100%
Overloading of Cargo Tank* 1.2E-­‐04 0% 100%
Source: Brandsæter and Hoffman (2010 p. 5-­‐75) and DNV (2000) as cited in Brandsæter and Hoffman (2010 p. 6-­‐92).
Note: Spill Probabilities represent loading and discharge for every event type with the exception of overloading of cargo tank.
* Overloading of cargo tank can only occur during loading operations.
DNV presents unmitigated risk as return periods for spills, which it calculates based on
forecast tanker traffic calls to Kitimat Terminal and spill size distributions for
particular types of cargo transfer operations. DNV estimates that unmitigated small or
3 According to Brandsæter and Hoffman (2010 p. 6-­‐91), the International Tanker Owners Pollution
Federation Limited categorizes spills as follows: small spills (700 tonnes ~ 1,000m 3).
10

medium size oil/condensate spills from cargo transfer operations will occur once every
29 years (see Table 10).
DNV applies mitigation measures to its unmitigated risk evaluation and recalculates
return periods for spills at the marine terminal. DNV quantitatively evaluates a closed
loading system with vapour recovery as the principal mitigation measure for spills at
the marine terminal. DNV claims that the closed loading system and vapour recovery
unit will eliminate any risk of an oil spill associated with overloading cargo tanks and
thus the probability per year of this event occurring drops to zero (Brandsæter and
Hoffman 2010 p. 131). According to DNV, mitigation measures more than double the
return period for unmitigated oil/condensate spills from 29 to 62 years (Table 10).
Table 10: Comparison of Overall Unmitigated and Mitigated Oil/Condensate Spill Return Periods
for Cargo Transfer Operations
Cargo Transfer Operation
Unmitigated Return
Period (in years)
Mitigated Return
Period (in years)
Release from Loading Arm 89 89
Equipment Failure 891 891
Failure in Vessels Piping System or Pumps 631 631
Human Failure 631 631
Mooring Failure 1,196 1,196
Overloading of Cargo Tank* 56 n/a
Total 29 62
Source: Brandsæter and Hoffman (2010).
The Bercha Group (2012a) prepared a risk analysis examining spills associated with
operations at the marine terminal. The risk assessment examines worker and public
safety at the marine terminal measured according to the risk to individuals exposed to
oil or condensate releases (Bercha Group 2012a). The Bercha Group examines failure
frequencies that include the following terminal components: piping, pumps, storage
tanks, and loading/unloading arms (Bercha Group 2012a). Based on calculations from
the Bercha Group report, the return period for a rupture from one of the terminal
components is 282 years (Table 11).
Table 11: Rupture Failure Frequencies for Marine Terminal Components
Component
Rupture Failure
Frequency
Estimated Return Period
(in years)*
Piping 1.00E-­‐07 10,000,000
Pumps 1.00E-­‐04 10,000
Oil Tanks 5.50E-­‐05 18,182
Condensate Tanks 1.50E-­‐05 66,667
Loading Arm -­‐ Oil 2.29E-­‐03 437
Unloading Arm -­‐ Condensate 1.09E-­‐03 917
Total*
Source: Bercha Group (2012a p. 4.4).
3.55E-­‐03 282
* We estimate return periods based on information and data from Bercha Group (2012a).
11

The Bercha Group (2012a) concludes that risks to individuals from operations at the
marine terminal are within tolerable limits. Based on Individual Specific Risks
thresholds, the authors determine that risks to the public in the area surrounding the
terminal are insignificant at less than one in one million, while risks to workers are
higher although tolerable at a range of between one in 10,000 and one in 100,000. The
authors recommend mitigation measures to ensure safe operational and work
procedures at the marine terminal.
3.3. Spills from Pipeline Operations
The ENGP regulatory application contains several pipeline spill estimates. In its
regulatory application submitted in 2010, Enbridge prepared a section in Volume 7B
estimating return periods for spills that occur along the pipeline route. In 2012,
WorleyParsons and the Bercha Group submitted additional risk analyses estimating
pipeline spill likelihood for the project.
Volume 7B of the Enbridge application identifies return periods for a hydrocarbon 4
release that occurs along the pipeline route. Enbridge calculates spill return periods
with liquid pipeline failure frequency data from 1991 to 2009 from the National Energy
Board (NEB). Enbridge notes that NEB data include pipelines up to 50 years old that
use older technology and building material standards associated with an increased
frequency of failures compared to modern pipelines (Enbridge 2010b Vol. 7B, p. 3-­‐1).
Accordingly, Enbridge factored improved design features and mitigation planning
characteristic of modern pipelines into its failure frequency calculations and expects
improvement factors to decrease the probability of a pipeline failure relative to NEB
failure frequency data (Enbridge 2010b Vol. 7B, p. 3-­‐1).
Enbridge estimates return periods for various spill size categories of hydrocarbon
releases for six regions along the pipeline route. Enbridge bases spill size categories on
NEB classifications whereby a small spill is less than 30 m 3 , a medium spill ranges
between 30 and 1,000 m 3 , and a large spill exceeds 1,000 m 3 (Enbridge 2010b Vol. 7B).
Enbridge identifies the six physiographic regions along the pipeline route as the Eastern
Alberta Plains, the Southern Alberta Uplands, Alberta Plateau, Rocky Mountains,
Interior Plateau, and the Coast Mountains. Table 12 presents spill return periods
calculated by Enbridge for medium and large spills in the six regions. Although
Enbridge presents spill return periods separately for each region, we combine return
periods for medium and large spills for the entire length of the pipeline that results in a
return period of 28 years.
4 Enbridge (2010b Vol. 7B p. 3-­‐1) refers to pipeline spills as hydrocarbon spills. We interpret hydrocarbon
spills to include both oil and condensate and thus assume spill return periods represent oil or condensate
spills.
12

Table 12: Return Periods for Spills from the Northern Gateway Pipeline
Physiographic Region
Approximate Pipeline
Length (km)
Spill Return Period (in years)
Medium Large
Eastern Alberta Plains 166 287 669
Southern Alberta Plains 350 136 317
Alberta Plateau 44 1,082 2,525
Rocky Mountains 103 462 1,079
Interior Plateau 404 118 275
Coast Mountains 105 454 1,058
Total* 1,172 41 95
Combined Medium and Large Spills**
Source: Based on Enbridge (2010b Vol. 7B p. 3-­‐2).
28
* We combine spill return periods for each segment to represent the spill return period for the entire pipeline route.
** We combine return periods for medium and large spills into a single return period based on information provided in Volume 7B
of the ENGP regulatory application.
Note: Medium spills represent spills from 30 to 1,000 m3 and large spills represent spills > 1,000 m3. Enbridge contracted WorleyParsons to prepare a risk assessment entitled Semi-­‐
Quantitative Risk Assessment to determine spill frequencies for leaks and full-­‐bore
pipeline ruptures over the length of the ENGP pipeline 5 . The analysis uses several
methodological approaches and datasets including a failure frequency model that uses
data from recently constructed pipelines, particularly Enbridge Line 4, as well as
pipeline spill data from the US Department of Transportation Pipeline and Hazardous
Materials Safety Administration (PHMSA) from 2002 to 2009 (WorleyParsons 2012).
The WorleyParsons report identifies eight pipeline hazards and determines that the
probability of a 594-­‐bbl (94 m 3 ) oil pipeline leak is 0.249, which results in a return
period of 4 years (Table 13). The WorleyParsons report estimates an annual probability
of 0.0042 (return period of 239 years) for a full-­‐bore pipeline rupture releasing 14,099
bbl (2,238 m 3 ) of oil. In addition to estimating oil spills, WorleyParsons estimates
return periods for a 593 bbl (94 m 3 ) condensate leak of 4 years and a 5,183 bbl (823
m 3 ) condensate rupture of 273 years.
5 The WorleyParsons (2012) report focuses on full-­‐bore pipeline ruptures, which according to
WorleyParsons, is a pipeline spill that releases hydrocarbons in an unconstrained manner (WorleyParsons
2012 p. 4). The second attachment in the report contains a failure likelihood assessment by Dynamic Risk
Assessment Systems that estimates likelihoods for pipeline leaks. For simplicity, we refer to the
WorleyParsons report in its entirety including all the report’s attachments.
13

Table 13: Return Periods for Leaks and Full-­‐bore Ruptures
Return Period (in years)
Pipeline Hazard
Oil Pipeline Condensate Pipeline
Leak Rupture Leak Rupture
Internal corrosion 0 0 0 0
External corrosion 0 0 0 0
Materials and manufacturing defects 652 652 652 652
Construction defects 39 0 39 0
Equipment failure 5 0 5 0
Incorrect operations 47 0 47 0
Damage from third parties 436 1,308 1,385 4,149
Geotechnical and hydrological threats 0 530 0 530
Total 4 239 4 273
Source: WorleyParsons (2012); WorleyParsons (2012 as cited in WrightMansell 2012)
The Bercha Group prepared two risk assessments examining separate risks from pump
stations and pipeline ruptures associated with the ENGP pipeline. Both studies use a
similar methodological approach that includes determining hazards, assessing the
frequency of occurrence of a spill, evaluating spill consequences, assessing risk, and
identifying mitigation measures that reduce risk (Bercha Group 2012b; 1012c).
The first Bercha Group (2012c) study assesses worker and public safety risks from an
oil or condensate spill at Smoky River Station, the largest pump station along the
pipeline route in Alberta near Grande Prairie. Table 14 shows the failure frequencies
per year for combined piping and pump spills at the Smoky River Station.
Table 14: Frequencies for Piping and Pump Spills at Smoky River Station
Type of Spill
Failure Frequency
(per year)
Return Period
(in years)*
Release Volume
(m3) Oil Leak 3.01E-­‐03 332 50
Oil Rupture 6.02E-­‐04 1,661 1,000
Condensate Leak 1.22E-­‐03 820 50
Condensate Rupture
Source: Bercha Group (2012c).
2.45E-­‐04 4,082 300
* Calculated from Bercha Group (2012c).
The second Bercha Group (2012b) study identifies areas along the pipeline route that
are known to have high public concentrations and determines the risk to individuals of
a pipeline spill at those locations. The authors identify three heavily populated areas
that could be affected by a pipeline spill: A casino near Whitecourt, AB; Burns Lake, BC;
and Kitimat, BC.
To determine public risk from a pipeline spill, the Bercha Group (2012b) estimates the
failure frequency of the ENGP pipeline based on historical pipeline rupture data from
the NEB from 1991 to 2009 and makes several adjustments to the data to incorporate
various improvements in pipeline operations. As shown in Table 15, the failure rate of
0.372 per 10,000 km per year (10,000 km-­‐year) based on NEB historical data
decreases significantly to 0.109 spills per 10,000 km-­‐year due to numerous downward
14

adjustments ranging from 60 to 90%. The Bercha Group (2012b) concludes that risks
to public safety from a pipeline failure near the Casino at Whitecourt, Burns Lake, and
Kitimat are insignificant. Risks to residents and indoor and outdoor workers from a
pipeline releasing oil, condensate, or both are all below one in one million per year,
which the Bercha Group determines are acceptable levels of risk (Bercha Group
2012b).
Table 15: NEB Pipeline Failure Frequencies Adjusted Downward to Represent the ENGP
Pipeline
Failure
Causes
NEB Failure
Rate
(per 10,000
km-­‐year)
ENGP
Failure Rate
(per 10,000
km-­‐year)
Metal Loss 0.085 0.017 80%
Cracking 0.128 0.013 90%
External
Interference
Geotechnical
Failure
Material
Failure
0.021 0.021 -­‐ n/a
0.021 0.021 -­‐ n/a
0.021 0.009 60%
Other Causes 0.096 0.028 71%
Total 0.372 0.109 71%
Source: Bercha Group (2012b).
Downward
Adjustment Reason for Adjustment
Modern piggable pipelines are less
likely to fail than pipelines
represented in NEB data
Advanced coating technologies, low
stress design, and fatigue avoidance
through design and operations
Construction quality control,
inspection/testing after construction,
and operational surveillance
Estimated as same percentage as NEB
data (approx. 26%)
3.4. Summary of Tanker, Terminal, and Pipeline Spill Estimates for the ENGP
Table 16 presents a brief summary of spill return periods from the ENGP regulatory
application. For tanker spills, return periods range between 78 years for an
unmitigated oil or condensate spill of any size to 15,000 years for a mitigated oil spill
exceeding 40,000 m 3 . Terminal spills range from one unmitigated oil/condensate spill
of any size occurring every 29 years to one mitigated spill every 430 years for medium
oil spills releasing 10 to 1,000 m 3 of oil. Pipeline spills range from a leak releasing 94
m 3 every 2 years to a full-­‐bore rupture releasing 2,238 m 3 of oil every 239 years. The
ENGP regulatory application does not evaluate overall spill likelihood for tanker,
terminal, and pipeline spills. If spill return periods are evaluated collectively for the
entire ENGP, data from the ENGP application suggest that a tanker or terminal or
pipeline spill could occur once every 2 years 6 .
6 Return period calculated based on the following spill probabilities: (1) any size tanker spill; (2) any size
terminal spill, and; (3) pipeline leak.
15

Table 16: Summary of Spill Return Periods for the ENGP
Type of Spill
Tanker Spill
Return Period
(in years)
Any size spill
Oil/Condensate
Oil
78 -­‐ 250
110 -­‐ 350
> 5,000 m
Oil spill exceeding a certain size
3 200 -­‐ 550
> 20,000 m3 1,750 -­‐ 2,800
> 40,000 m3 Terminal Spill
12,000 -­‐ 15,000
Any size spill
Oil/Condensate
Oil
29 -­‐ 62
34 -­‐ 90
Small spill (< 10 m3) Oil/Condensate
Oil
77
110
Medium spill (10 -­‐ 1,000 m3) Pipeline Spill
Oil/Condensate
Oil
46 -­‐ 294
49 -­‐ 430
Leak (94 m3) Oil/Condensate
Oil
2
4
Rupture # Tanker, Terminal, or Pipeline Spill
Oil/Condensate
Oil
128
239
Spill from ENGP*
Oil/Condensate
Oil
2
4
Source: Brandsæter and Hoffman (2010); Worley Parsons (2012).
Note: Range represents mitigated and unmitigated spill probabilities with the exception of pipeline spills
* Overall spill probability for ENGP based on spill probabilities for any size tanker spill, any size terminal spill, and pipeline leaks.
# Average size oil pipeline rupture is 2,238 m3 and average size condensate rupture is 823 m3 (WorleyParsons 2012).
4. Evaluation of ENGP Spill Estimates
The following section evaluates spill return periods presented in the ENGP regulatory
application. The objective of the assessment is to examine whether the methodological
approach used by Enbridge and its consultants adequately assesses the likelihood of
significant adverse environmental effects as required in the CEAA. To achieve this
objective, we assess methodologies estimating return periods for tanker, terminal, and
pipeline spills in the ENGP application with best practice criteria and identify any
deficiencies.
We examined several sets of guidelines for risk assessment practices and principles
and synthesized the literature into a single evaluative framework (Table 17). We use the
following four-­‐point scale to assess the degree to which each criterion is met:
• Fully met: no deficiencies
• Largely met: no major deficiencies
• Partially met: one major deficiency
• Not met: two or more major deficiencies.
16

Table 17: Evaluative Criteria for Assessing the Quality of Methodological Approaches to
Estimating Risk
Criterion Description Source
Transparency
Replicability
Clarity
Reasonableness
Reliability
Validity
Risk
Acceptability
Documentation fully and effectively discloses supporting
evidence, assumptions, data gaps and limitations, as well as
uncertainty in data and assumptions, and their resulting
potential implications to risk
Documentation provides sufficient information to allow
individuals other than those who did the original analysis to
obtain similar results
Risk estimates are easy to understand and effectively
communicate the nature and magnitude of the risk in a manner
that is complete, informative, and useful in decision making
The analytical approach ensures quality, integrity, and
objectivity, and meets high scientific standards in terms of
analytical methods, data, assumptions, logic, and judgment
Appropriate analytical methods explicitly describe and
evaluate sources of uncertainty and variability that affect risk,
and estimate the magnitudes of uncertainties and their effects
on estimates of risk
Independent third-­‐party experts review and validate findings
of the risk analysis to ensure credibility, quality, and integrity
of the analysis
Stakeholders identify and agree on acceptable levels of risk in a
participatory, open decision-­‐making process and assess risk
relative to alternative means of meeting project objectives
BC MoELP (2000);
US EPA (1998); US
EPA (2000); US EPA
(2004); IALA (2008)
US EPA (1998);
IALA (2008); NRC
(1996)
US EPA (1998); US
EPA (2000); US EPA
(2004)
BC MoELP (2000);
US EPA (1998); US
EPA (2000); NRC
(1996)
BC MoELP (2000);
US EPA (2000); US
EPA (2004); NRC
(1996)
US EPA (2000); US
EPA (2004); IALA
(2008)
BC MoELP (2000);
IALA (2008); NRC
(1996)
Transparency
Criterion: Documentation fully and effectively discloses supporting evidence, assumptions,
data gaps and limitations, as well as uncertainty in data and assumptions, and their
resulting potential implications to risk.
There are six major deficiencies related to transparency in the methods estimating spill
return periods in the ENGP regulatory application. Major deficiencies include:
1. Lack of supporting evidence for LRFP tanker incident frequency data
The frequency assessment in chapter five of the Marine Shipping Quantitative Risk
Analysis and described in section 3.1, which is the foundation for estimating return
periods for tanker spills, fails to provide the necessary supporting evidence.
First, there is inadequate information describing the LRFP data that contains
grounding, collision, foundering, and fire/explosion incidents. DNV fails to provide
proprietary data in an appendix or data file that includes information on the type
of accident, location, date, factors contributing to the incident, and whether or not
the incident resulted in a spill. Instead, DNV simply states that LRFP data are
obtained from 1990 to 2006 “…since the type of vessels in operation and the
incidents that have occurred after 1990 are considered to be more representative
of modern tanker operations, such as the one planned by Northern Gateway.”
17

(Brandsæter and Hoffman 2010 p. 5-­‐49). This lack of transparency raises concerns
related to the number and types of tanker incidents included in the data set and
whether the analysts altered any of the LRFP data based on undisclosed
assumptions. Further, LRFP data likely reflects jurisdictions where marine safety
measures such as those announced by the Canadian government and mitigation
measures such as escort tugs are already enforced. Thus incident frequencies
potentially double count the use of escort tugs that DNV incorporates into LRFP
data to reduce the likelihood of tanker spills although it is impossible to determine
without access to the proprietary data (potential double-­‐counting of mitigation
measures discussed on p. 33).
Second, DNV fails to support assumptions used in the development of tanker
incident frequencies. DNV claims that assumptions used in the development of
base incident frequencies per nm are based on information from tanker operators
and captains, as well as studies of vessel operating patterns (Brandsæter and
Hoffman 2010 p. 5-­‐50) yet there is no evidence provided to support these
assumptions. Two of the four assumptions used to estimate the average distance
travelled by a tanker per year are supported with reference to a study completed
for the liquefied natural gas terminal Rabaska (Rabaska 2004 as cited in
Brandsæter and Hoffman 2010 p. 5-­‐51). However, detailed information or
discussion comparing the similarities and differences between Rabaska and the
ENGP is absent from the report and the Rabaska study is not appended to DNV’s
study nor is it found in the project’s public registry database on the NEB website.
Furthermore, assumptions related to the amount of time tankers sail in areas
where a grounding could occur and the amount of time tankers sail in open water
where foundering can occur are not supported with any evidence or references nor
are any of the assumptions calibrated with historical data or expert opinion.
Incident frequencies are the basis for estimating spill return periods in the ENGP
application. Since return periods are the product of incident frequencies,
conditional probabilities, the distribution of tanker routes travelled, the length of
each tanker route, and mitigation measures, any uncertainty or errors in incident
frequencies will carry through to the final result. Thus, given the importance of
incident frequencies as a critical data input, DNV must effectively disclose all
adjustments, assumptions, and uncertainties in a transparent manner. The lack of
evidence in the ENGP regulatory application makes it impossible to assess the
validity of these values.
2. Insufficient evidence supporting conditional probabilities for tanker spills
The consequence assessment portion of the marine shipping QRA fails to
adequately disclose any supporting information used to estimate conditional
probabilities that an incident will result in a spill. DNV uses two different methods
to estimate conditional spill probabilities: the first method determines conditional
spill probabilities based on LFRP data; the second method estimates spill
quantities for bottom and side damages for groundings and collisions based on the
International Marine Organization International Convention for the Prevention of
18

Pollution from Ships (MARPOL) and Naval Architecture Package software. In the
first method, DNV simply claims that conditional spill probabilities are “based on
the research of spill to damage data” (Brandsæter and Hoffman 2010 p. 6-­‐79)
without providing the dataset from which it estimates conditional spill
probabilities. Similarly, DNV does not support conditional probability estimates
for various spill volumes in the second method with proprietary data and fails to
provide adequate information describing the nature of the original data used in the
analysis such as its assumptions, data gaps, and limitations. Similar to incident
frequencies, the importance of conditional probabilities as a major data input into
DNV’s methodological approach requires effective transparency of the data and
methods used to calculate conditional probabilities. Yet again, a lack of
transparency prevents validation of these important conditional probabilities.
3. Limited transparency in the application of local scaling factors to the BC study area
DNV completed a qualitative assessment to develop local scaling factors for tanker
traffic navigating the BC Coast that contains several deficiencies. The information
gathering process to determine scaling factors incorporated data obtained from
local meetings and interviews with tug boat operators, barge operators, sports
fishermen, environmental groups, and terminal operators (Brandsæter and
Hoffman 2010 p. 4-­‐46). After scaling factors were developed, a workshop held by
DNV and attended by experts in marine risk assessment, tanker operations, and
navigation validated the findings (Brandsæter and Hoffman 2010 p. 5-­‐52).
Despite the inclusive approach and the effort to engage local stakeholders, there is
a lack of evidence and transparency with how certain judgments for scaling factors
were made. DNV assesses local scaling factors against a baseline factor, yet does
not provide information on comparative areas used to determine the scaling factor
nor does it provide adequate evidence supporting the assigned scaling factor. An
example illustrates the subjectivity of scaling factors used to localize incident
frequencies. For powered grounding accidents, one of the three scaling factors,
navigational route, represents the number of course changes and the distance to
shore. The ‘world average’ of 1.0 for the scaling factor navigational route describes
areas where the “distance to shore or shallow water is approximately 4 nm, and
with very few critical course changes” (Brandsæter and Hoffman 2010 p. 5-­‐54).
Based on this ‘world average’, scaling factors for each segment range from 0.001 to
2.1 and ambiguous statements such as “narrow channel and consecutive course
changes” or “some navigational challenges and course changes” support scaling
factors (Brandsæter and Hoffman 2010 p. 5-­‐55). What exactly is meant by
“consecutive course changes” or “some navigational challenges”? What is the
difference between assigning a 1.5 for one segment compared to assigning a 1.8 for
another segment and what differences in conditions do these different scaling
factors describe? Is there a mathematical relationship between scaling factors for
different segments? Was there a methodical approach to assigning these values
and if so, what were the criteria? What navigational routes are included in the
world average? Similar concerns related to subjectivity and ambiguity exist for
19

other scaling factors that adjust powered grounding, drift grounding, collision,
foundering, and incidents involving fire or explosion to the BC context.
Furthermore, DNV does not clearly identify whether it assesses scaling factors
based on the comparison of the ENGP route to all other routes in the world for
which the data is collected or whether it assesses factors for the routes that are
close to shore in an area similar to the BC study area. Each approach would
produce significantly different estimates that result in material changes to incident
frequencies. Consequently, there is no basis on which to assess the validity of the
scaling factors.
4. Lack of information provided to compare incident frequencies at the proposed
Kitimat Terminal to marine terminals in Norway
DNV does not provide adequate evidence to support its claim that “Incident
frequencies from terminals in Norway are most representative of the operation
planned for the Kitimat Terminal and should provide an appropriate forecast of the
possible incident frequency at the Kitimat Terminal” (Brandsæter and Hoffman
2010 p. 5-­‐71). Given that the organization, DNV Maritime and Oil & Gas, is
headquartered in Norway, it seems appropriate that DNV use proprietary research
from its country of origin. However, DNV fails to make any comparison of marine
terminals in Norway to the proposed terminal in Kitimat. What are the geographic,
physical, biological, and socioeconomic similarities between terminals in both
regions? What are the differences in regulatory environments and marine safety
practices between Canada and Norway? How do historical incident frequencies
compare at terminals in Canada and Norway? These questions represent only
some of the comparisons that should be made to support DNV’s decision to use
incident frequencies from Norway to forecast possible incident frequencies at
Kitimat Terminal.
5. Lack of transparency for mitigation measures that reduce the likelihood of tanker
and terminal spills
The fifth major deficiency concerning transparency relates to mitigation measures
and their risk reducing effect on spill return periods. In Chapter 8 of the Marine
Shipping Quantitative Risk Analysis, DNV examines risk reduction measures for
tanker and terminal operations. The evaluation largely consists of a qualitative
assessment of the characteristics of tug operations, enhanced navigational aids,
and vessel traffic management system, among others. The only quantitative risk
reduction factors considered in the methodological approach estimating spill
return periods are escort tugs for tankers and a closed loading system at the
marine terminal for cargo transfer operations. DNV estimates that both mitigation
measures significantly decrease spill return periods.
DNV bases the risk reducing effect of escort and tethered tugs for tanker incidents
on a confidential study it completed in 2002 for tug escort operations at Fawley
Terminal in the United Kingdom. According to this study, DNV claims that the use
20

of escort and tethered tugs will reduce the frequency of powered and drift
grounding events 80% to 90% and will reduce collision incidents 5% (Brandsæter
and Hoffman 2010 p. 8-­‐120). DNV determines that the use of escort and tethered
tugs will increase return periods from 78 years to 250 years for any size oil or
condensate spill thus producing a three-­‐fold risk reduction effect compared to
unmitigated return periods. Given the significant increase in return periods
forecast to occur with the use of escort and tethered tugs for tanker traffic, it is
important for DNV to transparently provide quantitative analysis documenting the
impact of mitigation. DNV must disclose specific information related to the
number of observations on the role of these mitigation measures and the reliability
of the data set regarding their impact from the confidential study DNV prepared in
2002 that is the foundation for reductions from mitigation measures in the marine
shipping QRA.
Similar concerns exist for mitigation measures at the marine terminal. In its
limited assessment of mitigation measures for terminal operations, DNV claims
that the closed loading system installed at Kitimat Terminal will eliminate the
likelihood of spills from overfilling cargo tanks. Even in the event cargo tanks are
overfilled, the closed loading system would redirect oil into nearby empty ship
tanks “thereby eliminating the risk of an oil spill” (Brandsæter and Hoffman 2010
p. 131). DNV does not provide any historical evidence or data on the performance
of closed loading systems to support this statement, which is concerning given that
this single assumption eliminates any risk of a cargo tank overloading and more
than doubles the unmitigated spill return period of 29 years to a mitigated return
period of 62 years. Furthermore, given that the average loading/discharge
operations at Kitimat Terminal is approximately 30 hours per year for each tanker
(Brandsæter and Hoffman 2010 p. 5-­‐74), there is no guarantee that a nearby
tanker will be berthed at Kitimat Terminal in which to redirect oil into empty ship
tanks.
6. Inadequate evidence supporting the reduction of spill frequencies for pipeline
operations
The methodological approaches in the ENGP application estimating return periods
for pipeline spills in Volume 7B and pipeline failure frequencies that affect public
safety in the Bercha Group (2012b) study lack transparency. Both studies
incorporate downward adjustments into calculations for pipeline spill frequencies
without providing adequate evidence.
Enbridge estimates return periods for a pipeline spill in Volume 7B by adjusting
spill frequency data from the NEB without disclosing how it made the downward
adjustments and fails to provide evidence to justify the reductions. According to
Enbridge (2010b Vol. 7B p. 3-­‐1), NEB data from 1991 to 2009 7 includes pipelines
up to 50 years old that use older technology and standards of building materials,
7 Note that there is a discrepancy on page 3-­‐1 of Volume 7B, since Enbridge references NEB data from 1991 to
2000 as well as data from 1991 to 2009.
21

are not amenable to internal inspection, contain material that did not undergo
inspection or quality control, and may have other characteristics associated with
more failures compared to modern pipelines. Based on these assumptions and
improved design and mitigation planning associated with modern pipelines,
Enbridge claims that the aforementioned factors decrease the likelihood of
pipeline failure and “were factored into failure frequency calculations for the
pipelines” (Enbridge 2010b Vol. 7b p. 3-­‐1). Enbridge does not provide proprietary
data from the NEB, fails to present any spill rates or spill frequencies calculated
from the NEB data, fails to explicitly explain how it adjusted incident frequencies,
and presents no evidence to justify the downward adjustments. Furthermore, it is
unclear whether the analysis includes the crude oil pipeline, condensate pipeline,
or both pipelines combined.
Similarly, the Bercha Group (2012b) estimates failure frequencies for the ENGP
pipeline based on historical pipeline rupture data from the NEB from 1991 to 2009
and makes several adjustments to the NEB data to incorporate various
improvements in pipeline operations. The authors fail to adequately justify
downward adjustments to several causes of pipeline ruptures including metal loss,
cracking, and material failure that result in a decrease in the NEB failure rate of
71%. First, the Bercha Group reduces the failure frequency for metal loss (internal
and external corrosion) by 80% based on a study by Muhlbauer (1996) entitled
Pipeline Risk Management Manual and discussions with experts, although the
authors provide no evidence to justify applying the reduction factor to the ENGP
pipeline system other than suggesting that the ENGP pipeline is modern. Second,
the Bercha Group reduces the failure frequency for cracking for the ENGP pipeline
system by 90% compared to the NEB data claiming that “advanced coating
technologies, low stress design, no tape coat, fatigue avoidance through design and
operations, is expected to virtually eliminate cracking potential” (Bercha Group
2012b p. 3.5) without referencing a single study, expert opinion, data set, or other
source of evidence. Third, the authors reduce the pipeline rupture frequency for
material failure by 60% due to quality control during construction, inspection,
testing, and other factors. Once again the report cites Muhlbauer (1996) but fails
to justify the application of the reduction factor to the ENGP pipeline system and
lacks evidence describing how the mitigation measures will achieve the stated
reductions in spill frequency.
Historical spill performance for Enbridge’s own pipeline system shows an increase
in spill frequency. According to data submitted by Enbridge (2012a) to the Joint
Review Panel for the ENGP providing the number of reportable spills per thousand
km of liquids pipeline, the frequency of spills increases from 1998 to 2010 (see
linear trendline in Figure 2). We recognize that this period is an insufficient amount
of time to evaluate historical spill data but Enbridge does not publicly provide
comprehensive spill data prior to 1998. We also recognize that spill data are not
broken down by age of pipeline to allow for a direct comparison of newer to older
pipelines. Nonetheless, Enbridge’s own data show an increase in spill frequency
that contradicts the downward adjustments that reduce spill frequencies in the
22

ENGP regulatory application. Furthermore, a comprehensive study of spill rates in
the US also found that there was no decline in pipeline spill rates between 1972
and 2005 (Eschenbach et al. 2010) 8 .
Figure 2: Historical Spill Performance of the Enbridge Liquids Pipeline System (1998 -­‐ 2010)
Source: Enbridge (2012a)
Recent spill data for the Keystone pipeline illustrates that new pipelines remain
prone to spills. TransCanada operates the 3,460-­‐km Keystone pipeline that
delivers crude oil from Hardisty, AB to the US Midwest. As shown in Table 18, a total
of 12 spills occurred in the US leg of the pipeline in the first year the Keystone
began operation. Although the majority of the spills were less than 1 bbl, the 12
reported spills from the modern Keystone pipeline released a total of 505 to 555
bbl (80 to 88 m 3 ) in a single year of operation. Although the Keystone data is
limited, it raises questions about the assumption of improved safety of new
pipelines.
8 Based on data representing US oil production in the Outer Continental Shelf region of the Gulf of Mexico
from the Mineral Management Service between 1972 and 2005, Eschenbach et al. (2010) determine that
although the number of pipeline spills < 1,000 bbl declined from 16 in 1972-­‐1989 to 4 in 1990-­‐2005, the
decline does “… not demonstrate a statistically significant drop over time in the rate of pipeline spills” (p. 3-­‐
4).
23

Table 18: Recent Spill Record for the Keystone Pipeline (2010 -­‐ 2011)
Incident Date Cause of Incident
Volume of Oil Spilled
(bbl)
May 21, 2010 Valve body leak < 1
June 23, 2010 Pump station leak < 1
August 10, 2010 Unknown < 1
August 19, 2010 Equipment failure < 1
January 5, 2011 Faulty seal < 1
January 30, 2011 Pump seal failure < 1
February 3, 2011 Operator error < 1
February 17, 2011 Equipment failure < 1
March 8, 2011 Pump seal failure < 1
March 16, 2011 Seal failure 3
May 7, 2011 Valve failure 450 – 500
May 29, 2011 Unknown 50
Total 505 – 555
Source: USCG NRC (2012); USDS (2011 Vol. 2 p. 3.13-­‐13).
Evaluation: There are six major deficiencies related to the transparency criterion and
thus this criterion is not met.
Replicability
Replicability: Documentation provides sufficient information to allow individuals other
than those who did the original analysis to obtain similar results.
The ENGP regulatory application presents the methodologies used to estimate spill
return periods for tanker, terminal, and pipeline spills in a fairly straightforward
manner. However, insufficient information due to a lack of transparency prevents the
replication of key components of the methodological approach and important results in
the ENGP regulatory application. As suggested in the previous section, inadequate
transparency prevents individuals other than the original analysts from replicating the
following results and each component represents a major deficiency related to
replicability:
1. LRFP frequency data for grounding, collision, foundering, and fire/explosion tanker
incidents
2. Scaling factors for ENGP routes due to a failure to provide comparative navigational
routes used to determine scaling factors and methodology for determining scaling
factors
3. Conditional spill probabilities and spill size distributions estimated in the consequence
assessment for tanker spills
4. Mitigation measures that reduce spills from ENGP tanker traffic and marine terminal
operations
5. Incident frequencies and mitigation measures used to estimate return periods for
pipeline spills.
24

We acknowledge the difficulty in replicating certain results, particularly those based on
random sampling from methods such as Monte Carlo simulation. However, proprietary
data and the computer code for the Monte Carlo simulation model should be included
in an appendix or separate data report in order to allow an independent party to
conduct similar tests and compare results with those included in the ENGP regulatory
application.
Evaluation: There are five major deficiencies related to the replicability criterion and thus
this criterion is only not met.
Clarity
Criterion: Risk estimates are easy to understand and effectively communicate the nature
and magnitude of the risk in a manner that is complete, informative, and useful in
decision-­‐making.
Decision makers must use the criterion for likelihood of significant adverse
environmental effects for project assessment under CEAA. Methodologies that
calculate and present spill likelihood in the ENGP regulatory application do not provide
a clear assessment of the likelihood of spill occurrence. There are five major
deficiencies related to clarity in the methodological approach for estimating return
periods for ENGP tanker, terminal, and pipeline spills:
1. Failure to effectively communicate the probability of a spill over the life of the project
Enbridge expresses the likelihood of a spill as a return period rather than the
probability of a spill over the life of the project, which presents a major deficiency
in the communication of spill estimates to decision-­‐makers. A return period is the
number of years between spill events and is the inverse of the probability. Thus, if
the unmitigated return period for any size oil or condensate spill is 78 years, there
is a 1 in 78 chance, or approximately 1% probability, that a spill will occur each
year. Return periods incorrectly imply that an oil spill event will occur only once
throughout the recurrent interval, such as 78 years, when in fact the event can
occur numerous times or not at all. Furthermore, return periods do not
communicate the probability of a spill during the operational life of the ENGP,
which ranges from 30 to 50 years 9 . Without estimates of the probability of a spill
over the operating life of the project, decision makers are unable to determine
whether the project meets the CEAA criterion of likelihood of adverse
environmental effects. Indeed, stating that there is a 22.2% chance of a tanker spill
exceeding 5,000 m 3 over the operating life of the ENGP, instead of reporting that
9 The Wright Mansell report entitled Public Interest Benefits of the Enbridge Northern Gateway Pipeline Project
appended to Enbridge (2010b Vol. 2) assumes a 30-­‐year operating period whereas Enbridge assumes a 50-­‐
year operating period for pipeline operations in Volume 7B: Risk Assessment and Management of Spills –
Pipelines.
25

there will be one spill every 200 years, more effectively communicates the
likelihood of a spill to decision-­‐makers 10 .
2. Failure to combine estimates for tanker, terminal, and pipeline spills to present a
single spill estimate for the entire project
A second major deficiency is Enbridge’s failure to estimate spills for the entire
ENGP. The risk assessment for marine transportation in Volume 8C, the risk
assessment for pipelines in Volume 7B, the marine shipping QRA prepared by DNV,
and the Bercha Group (2012a; 2012b; 2012c) and WorleyParsons studies all
present separate spill estimates. Enbridge does not combine separate estimates
for tanker, terminal, and pipeline spills, as well as their component parts including
pumps, storage tanks, and pump stations to demonstrate the collective likelihood
of a spill from all potential spill sources. By presenting separate spill return
periods for individual components of the project instead of the entire project,
Enbridge underestimates spill likelihood and fails to provide decision makers with
the overall spill risk information necessary for applying the CEAA decision
criterion.
To address the lack of clarity in the ENGP application, we restate spill return
periods estimated by Enbridge and their consultants by converting return periods
to annual spill probabilities and estimate spill probabilities over the life of the
project rather than on an annual basis 11 . Based on Enbridge’s analysis, the
restated probability of any size oil or condensate tanker spill over the life of the
ENGP ranges from 11.3% to 47.5% (Table 19). The probability of a spill increases
significantly to 99.9% when spill probability is evaluated for the entire ENGP
inclusive of tanker, terminal and pipeline spills 12 . Our revised estimates in Table 19
likely underestimate spill probabilities because they are based on estimates in the
ENGP regulatory application that understate risk due to data limitations and
undocumented mitigation adjustments 13 . Therefore actual spill probabilities are
likely higher.
10 To estimate the chance of a 5,000 m 3 oil spill over a 50-­‐year operating period, we calculate the inverse of
the return period for a 5,000 m 3 spill (200 years) to estimate the annual probability of a spill (0.005 or 0.5%)
and use the formula in footnote 11 to obtain a 22.2% probability over 50 years.
11 We use the following formula to convert annual probabilities to probabilities over a 30-­‐ and 50-­‐year
period: 1 -­‐ ((1 -­‐ P) n), where P is the annual probability and n is the number of years.
12 Return period calculated based on the following spill probabilities: (1) any size tanker spill; (2) any size
terminal spill, and; (3) pipeline leaks. We note that Enbridge’s consultants estimated an overall probability of
70.9% over 50 years during the hearings (Enbridge 2012b Vol. 78 p. 53). However, this lower probability is
based on any size oil/condensate tanker spill (18.2%), any size oil/condensate terminal spill (56.2%), and an
oil pipeline rupture (18.8%). This approach is not representative of overall spill probability for the ENGP
because it relies on the likelihood of a pipeline rupture, which is less likely to occur than a leak, and omits
condensate pipeline spills.
13 Data limitations in the ENGP application include estimating spill likelihood for tankers with LRFP data that
potentially underreport actual tanker spills, omitting the majority of the distance travelled by tankers to
export markets, potential double-­‐counting of mitigation measures, and failing to consider characteristics of
the project that could increase risk such as the corrosiveness of diluted bitumen and maneuverability
26

Table 19: Spill Probabilities Over the Life of the ENGP Based on Regulatory Application
Probability over 30
Years (%)
Probability over 50
Years (%)
Tanker Spill
Any size oil/condensate 11.3 -­‐ 32.1 18.2 -­‐ 47.5
Any size oil 8.2 -­‐ 24.0 13.3 -­‐ 36.7
Oil > 5,000 m 3 5.3 -­‐ 14.0 8.7 -­‐ 22.2
Oil > 20,000 m 3 1.1 -­‐ 1.7 1.8 -­‐ 2.8
Oil > 40,000 m 3 0.2 0.3 -­‐ 0.4
Terminal Spill
Any size oil/condensate 38.6 -­‐ 65.1 55.6 -­‐ 82.7
Any size oil 28.5 -­‐ 59.2 42.8 -­‐ 77.5
Pipeline Spill^
Oil/Condensate leak 99.9 99.9
Oil leak 99.9 99.9
Tanker, Terminal, or Pipeline Spill*
Oil/Condensate 99.9 99.9
Oil 99.9 99.9
Source: Calculated based on Brandsæter and Hoffman (2010) and WorleyParsons (2012).
Note: Range represents mitigated and unmitigated spill probabilities with the exception of pipeline spills
* Overall spill probability for ENGP based on any size tanker spill, any size terminal spill, and pipeline leaks.
^ Average size oil or condensate pipeline leak is 94 m 3 (WorleyParsons 2012).
3. Failure to determine the likelihood of all spill size categories that can have a
significant adverse environmental impact
Another deficiency is the failure to estimate spill likelihood for all spill size
categories that can have a significant adverse environmental impact, particularly
smaller spills. In its marine shipping QRA, DNV presents return periods for any
size oil/condensate spill as well as return periods for several categories of oil spill
volumes including spills greater than 5,000 m 3 , 10,000 m 3 , 20,000 m 3 , and 40,000
m 3 . However, DNV fails to examine the likelihood of spills for specific spill volumes
less than 5,000 m 3 despite the potential significant effects of smaller spill volumes.
The US environmental impact assessment of potential oil spills in Cook Inlet,
Alaska (US DOI 2003a), an area with similar characteristics as the BC study area,
predicts that an oil spill of 238 m 3 (1,500 bbl) could have significant adverse
environmental impacts. The US DOI (2003a) environmental impact statement
assesses economic, environmental, and sociocultural impacts associated with the
exploration, development, and production of the Cook Inlet Outer Continental Shelf
in Alaska. One component of the study involves an assessment of the effects of a
large oil spill, which the US DOI defines as a spill greater than or equal to 1,000 bbl
(159 m 3 ). The authors use the OSRA model with spill rates from Anderson and
LaBelle (2000) to determine spill occurrence and an oil spill trajectory model that
reflects physical conditions of the environmental setting to determine the chance
that the spill might contact environmental resources. The authors then estimate
impacts to economic, environmental, and sociocultural resources affected by the
differences among tankers. In terms of undocumented mitigation adjustments, DNV does not provide
evidence for its downward adjustments that significantly reduce the likelihood of tanker and terminal spills.
27

hypothetical spill based on the modeling results, characteristics of the affected
environment, documented impacts from previous spills in the Alaska region (i.e.
Glacier Bay and Exxon Valdez) and peer-­‐reviewed studies. The Cook Inlet
environmental impact statement estimates the following potentially significant 14
impacts from a 1,500 bbl (238 m 3 ) oil platform spill or a 4,600 bbl (731 m 3 )
pipeline spill:
• Beach and intertidal fish habitat impacts could last more than a decade from
residual oil
• Mortality of endangered and threatened species such as sea otters and Steller
sea lions could occur
• Commercial fisheries could close for an entire season due to tainting
concerns
• Oil remaining in shoreline sediments for up to 10 years could impact clam
and shellfish sport fisheries
• Recreation and tourism areas could close partially or completely
• Oil spill cleanup activities could disturb archaeological resources
• Negative effects could occur to intrinsic values of National and State parks.
In addition to significant effects, the environmental impact statement also
identifies the following spill effects to other marine-­‐related resources that could
occur from a 238 to 731 m 3 oil spill in Cook Inlet:
• The spill could impact an area between 618 and 1,100 km 2
• Between 17 to 38 km of shoreline could be contaminated for up to a decade
• Water quality in the vicinity of the spill would be at chronic toxicity levels for
up to 30 days
• Local intertidal and lower trophic-­‐level organisms could be depressed
measurably for about one year
• Mortality of adult fish, fish fry, and eggs could occur, although there would be
no measurable loss to overall fish populations
• Mortality of hundreds to tens of thousands of birds could occur and recovery
could take from a few years to a few generations
• Mortality of small numbers of resident marine mammals and recovery could
take from one to five years, although no measurable decline in regional
populations are expected
14 The US DOI uses a definition for significance consistent with the Council on Environmental Quality National
Environmental Policy Act (NEPA) regulations, which defines significance in terms of context and intensity (US
DOI 2003a p. IV-­‐1):
“Context” considers the setting of the Proposed Action, what the affected resource might be, and whether the effect on
this resource would be local or more regional in extent. “Intensity” considers the severity of the impact, taking into
account such factors as whether the impact is beneficial or adverse; the uniqueness of the resource (for example,
threatened or endangered species); the cumulative aspects of the impact; and whether Federal, State, or local laws
may be violated…Our (US DOI) EIS (environmental impact statement) impact analyses address the significance of the
impacts on the resources considering such factors as the nature of the impact (for example, habitat disturbance or
mortality), the spatial extent (local and regional), temporal and recovery times (years, generations), and the effects of
mitigation (for example, implementation of the oil-­‐spill-­‐response plan).
28

• Mortality of a small number of terrestrial mammals and recovery could take
one to three years
• Disproportionately high adverse effects would occur to Native populations
resulting from potential contamination of subsistence harvest areas,
tainting concerns and disruption of subsistence practices.
In summary, DNV fails to determine the likelihood of a spill for all spill sizes that
can have a significant adverse environmental impact and fails to provide
supporting evidence for using 5,000 m 3 as the implied cut off for significant
adverse effects.
4. Lack of clear communication of spill estimates generated in detailed technical data
reports in the main ENGP application
A fourth deficiency is Enbridge’s failure to clearly state spill estimates generated in
the detailed technical reports in the main application report. For example,
Enbridge summarizes the 139-­‐page Technical Data Report Marine Shipping
Quantitative Risk Analysis prepared by DNV in just 23 pages in the TERMPOL Study,
in just two pages in Volume 7C, and four pages in Volume 8C of the ENGP regulatory
application. The executive summary of the Enbridge application summarizes the
risk of tanker spills in just one paragraph that states (Enbridge 2010b Vol. 1, p.11-­‐
35):
For risk related to vessel transits within the Territorial Sea of Canada with inclusion
of navigation safety mitigation (primarily the use of tethered and close escort tugs),
the calculated return period of a spill of any size from a tanker carrying oil is 350
years. For condensate, the return period is 890 years. The return period for large
scale releases increases (risk level decreases) substantially to a level of 550 years for
a spill exceeding 5,000 m 3, 2,800 years for a spill exceeding 20,000 m 3 and more than
15,000 years for a spill exceeding 40,000 m 3.
A significant amount of important information required to inform decision-­‐makers
is missing from both the TERMPOL study and Volumes 1, 7C and 8C of the ENGP
application, including information on the methodological approach, data gaps and
limitations, assumptions used throughout the analysis, any uncertainties
associated with results, and the sensitivity analysis.
Enbridge presents spill return periods for pipeline operations in three pages in
Volume 7B and there is no reference to a technical data report or appendix
containing a more comprehensive analysis of pipeline spills. Thus not only is the
information on pipeline spills inadequate in Volume 7B, but no data report exists to
clarify the methodology, data, assumptions, and limitations that clearly and
comprehensively explain how pipeline spill return periods were evaluated 15 . The
15 Note that although the WorleyParsons (2012) report examines a full-­‐bore pipeline rupture and the Bercha
Group (2012b; 2012c) reports examine pipeline spills at pump stations and a pipeline spill in densely
29

disconnect between data reports and the main ENGP regulatory application, as
well as the complete absence of a data report for pipeline spills, reduce
accessibility to critical information, present barriers to the effective
communication of spill likelihood to stakeholders, and could negatively impact the
decision-­‐making process.
5. Failure to clearly address the effective implementation of mitigation measures that
reduce risk
Another consideration related to clarity concerns the presentation of mitigation
measures that purportedly reduce risk without discussing the effective
implementation of risk management measures. In the marine shipping QRA, DNV
identifies mitigation measures that it claims will reduce the likelihood of a spill.
Enbridge’s consultants state “Should any of these requirements not be met the risk
reduction effect would decrease accordingly.” (Brandsæter and Hoffman 2010 p. 8-­‐
120). Ensuring effective implementation is the responsibility of the company and
the NEB, which is responsible for regulating companies’ activities. The
enforcement and monitoring record of the NEB raises serious doubts about the
effectiveness of implementing risk management. According to an audit performed
by the Commissioner of the Environment and Sustainable Development (2011 p.
10), nearly two-­‐thirds (64%) of the compliance verification files reviewed by the
NEB identified deficiencies and only 7% of those files provided evidence the NEB
followed up with companies to determine if deficiencies were corrected. Further,
100% of the emergency response plans reviewed had deficiencies and there was a
follow-­‐up to address the deficiencies in only one case (CESD 2011 p. 11). Thus, if
Enbridge fails to implement mitigation measures and the NEB fails to take
corrective action, spill likelihood could significantly exceed Enbridge’ s estimates
that assume effective implementation of all risk-­‐reducing mitigation measures.
Furthermore, inadequate monitoring and enforcement have implications for
emergency response in the event that a spill from the ENGP occurs. If the NEB fails
to enforce emergency response initiatives outlined by Enbridge in the ENGP
application to respond to a spill, a similar situation may result as the 20,000-­‐bbl
diluted bitumen spill that shutdown sections of the Kalamazoo River near
Marshall, Michigan in July 2010 (NTSB 2012). Factual reports from the NTSB
investigation into the spill characterize Enbridge’s response as incompetent in the
detection and shutdown of the spill and suggest that Enbridge demonstrated
inadequate preparation to contain the spill. Indeed, Enbridge control room
operators failed to detect or attempt to shutdown the ruptured pipeline for 17
hours even though monitoring systems repeatedly sounded alarms and displayed
low-­‐pressure readings (NTSB 2012). After the spill was detected, Enbridge
experienced considerable difficulties locating contractors and other resources in
the region to contain the spill and waited for the US Environmental Protection
Agency to eventually provide the contact information for contractors (NTSB 2012).
populated areas, both of these reports are standalone studies that are independent from Volume 7B and were
submitted two years after the initial application.
30

In its regulatory application for the ENGP, Enbridge outlines a general spill
response plan and claims that equipment will be pre-­‐staged to improve response
time (Enbridge 2010b Vol. 8C p. 5-­‐1). However, the weak monitoring record of the
NEB (CESD 2011) and Enbridge’s recent mismanagement of the Line B spill in
Michigan suggest that any mitigation measures identified in the ENGP application
to reduce spill likelihood and minimize spill damage must include a risk
assessment of the likelihood of implementation failure and must be accompanied
by a detailed implementation plan that clearly outlines a comprehensive
monitoring and verification program 16 .
Evaluation: There are five major deficiencies related to the clarity criterion and thus this
criterion is not met.
Reasonableness
Reasonableness: The analytical approach ensures quality, integrity, and objectivity, and
meets high scientific standards in terms of analytical methods, data, assumptions, logic,
and judgment.
The methodological approach estimating spill return periods for the ENGP contains
five major deficiencies related to the reasonableness criterion. Major deficiencies
include:
1. Inappropriate definition of the study area to estimate spill return periods for tanker
operations
DNV calculates return periods for tanker spills based on an inappropriate
definition of the study area. Although a sensitivity analysis extends the transit area
to 200 nm off the BC coast, DNV does not calculate the likelihood of a spill for the
entire tanker voyage to any key export markets in Asia or the US, nor does it
examine the potential effects of a spill outside of the narrowly defined BC study
region.
DNV limits its analysis of tanker incident frequencies to three shipping routes
within Canada’s Territorial Sea that represent an average of 233 nm. Yet, one-­‐way
sailing distances from Kitimat, BC to the three key Northeast Asia markets of
Shanghai, China, Yokohama, Japan, and Ulsan, South Korea range from 4,041 to
4,865 nm (Enbridge 2010b Vol. 2 App. A p. 13) and the approximate one-­‐way
distance from Kitimat to San Francisco and Los Angeles, California are respectively
1,051 nm and 1,391 nm (Sea Distances -­‐ Voyage Calculator undated). Thus, DNV
ignores an average of nearly 4,200 nm to key markets in Asia and 1,000 nm to
markets in the US per tanker route, which represents a significant exclusion of the
distances navigated by tankers (Table 20). Assuming 149 oil tankers sail at least
one leg of their voyage to Asia or the US, one-­‐way distance shortfalls per tanker
16 We acknowledge that the Enbridge spill in Michigan occurred in a different jurisdiction than that regulated
by the NEB. However, we reference the mismanagement of the spill to demonstrate that poor
implementation of a spill response plan can have serious consequences.
31

amount to the omission of over 624,000 nm to Asia markets and over 147,000 nm
to markets in the US per year.
Table 20: Differences in Nautical Miles used to Estimate Tanker Spills and Actual Nautical
Miles to Key Export Markets
Key Export Market
One-­‐way
Distance to
Market
(in nm)
Average One-­‐
way Distance
in Study Area
(in nm)
One-­‐Way
Distance
Shortfall Per
Tanker
(in nm)
Total
Distance
Shortfall Per
Year
(in nm)
Asia Markets
Shanghai, China* 4,865
4,631 690,044
Yokohama, Japan*
Ulsan, South Korea*
4,041
4,363
233
3,808
4,129
567,342
615,246
Average
US Markets
4,423 4,189 624,211
San Francisco** 1,051
818 121,832
Los Angeles** 1,391 233
1,158 172,492
Average 1,221 988 147,162
Source: Brandsæter and Hoffman (2010 p. 3-­‐1); Enbridge (2010b Vol. 2 App. A p. 13); http://sea-­‐distances.com/
* We calculate nautical miles as half of the round-­‐trip distance to all three regions in Enbridge (2010b Vol. 2 App A. p. 13).
** We calculate nautical miles from Kitimat to San Francisco and Los Angeles using the calculator from http://sea-­‐
distances.com/
Failure to consider spills outside the study region is contrary to CEAA, which
requires an assessment of environmental effects that would occur “outside
Canada” (CEAA Sec. 5). Thus, DNV’s methodological approach that ignores the
majority of the distance sailed by tankers to key export markets significantly
underestimates spill likelihood and prohibits decision-­‐makers from determining
significant adverse environmental effects of the ENGP outside Canada required
under CEAA.
2. Failure to adjust tanker incident frequency data to correct for underreporting
Literature in peer-­‐reviewed sources suggests that vessel accident data reported in
the LRFP database, which DNV uses to determine tanker incident frequencies in
the marine shipping QRA, have serious deficiencies. Hassel et al. (2011) examine
the LRFP database for underreporting of foundering, fire/explosion, collision,
wrecked/stranded, contact with a pier, and hull/machinery accidents for merchant
vessels exceeding 100 gross tonnes registered in particular states (flag states)
including Canada, Denmark, Netherlands, Norway, Sweden, United Kingdom, and
the US from January 2005 to December 2009. Using various statistical methods,
the researchers estimate that reporting performance by LRFP ranges between 4%
and 62% for select flag states compared to actual accident occurrences, suggesting
that as few as one in 25 accidents were reported in the LRFP database for a
particular flag state over a five-­‐year period 17 . In the best-­‐case scenario for vessel
17 Hassel et al. (2011) suggest that results should be interpreted with caution due to assumptions and
estimation methods used by the authors. However, according to Hassel et al. (2011), their findings are
consistent with those made in the study by Psarros et al. (2010).
32

accidents in Canada, the LRFP database reports 69% of all accidents and thus omits
nearly one-­‐third (31%) of all accidents occurring for vessels with a Canadian flag
(Hassel et al. 2011). A separate study conducted by Psarros et al. (2010) observes
similar underreporting in the LRFP database for accidents from vessels registered
in Norway. Based on an analysis of accident data for merchant vessels exceeding
100 gross registered tonnage from February 1997 to February 2007, the
researchers estimate that at best only one in three (30%) occurred accidents are
reported in the LRFP database 18 . Thus, the LRFP database has no record for 70%
of accidents from vessels registered in Norway over a 10 year period.
Furthermore, Psarros et al. (2010) observe that the effect of the vessel’s size on
reporting performance is insignificant and that the seriousness of an accident does
not have a significant effect on the likelihood of the accident being reported. The
relationship between incident frequency underreporting in the LRFP database and
spill frequency is unknown due to DNV’s failure to include proprietary data from
LRFP in its study.
To address underreporting, Hassel et al. (2011) suggest that statistical accident
data should be accompanied by adjustments such as correction factors, safety
margins, or expert judgment. DNV did not adjust data derived from the LRFP
database to incorporate any uncertainties associated with LRFP data in their
methodological approach. Even if underreporting relates to minor incidents and
has only a minimal effect on incident frequencies, DNV should, at the very least,
disclose the known issue of incomplete LRFP data in a transparent manner and
describe why analysts did not adjust the data accordingly. DNV’s failure to
acknowledge deficiencies in LRFP data and to make adjustments to reflect
potential underreporting is particularly surprising given that the authors of the
article documenting underreporting were employees of DNV when the article was
published in 2010.
3. Potential double-­‐counting of mitigation measures for LRFP tanker incident data
DNV decreases the likelihood of tanker spills by adjusting LRFP data with
mitigation measures based on the use of escort and tethered tugs at Kitimat
Terminal. DNV claims that the use of escort and tethered tugs will reduce the
frequency of powered and drift groundings 80% to 90% and will reduce collision
incidents 5% compared to tanker incidents without these mitigation measures
(Brandsæter and Hoffman 2010 p. 8-­‐120). As a result of these mitigation
measures, spill likelihood decreases three-­‐fold from an unmitigated return period
of 78 years for any size oil or condensate tanker spill to a mitigated return period
of 250 years.
LRFP tanker incident data may already represent locations that use escort tugs for
tanker traffic and thus DNV’s downward adjustments to the LRFP data may
double-­‐count mitigation measures that reduce spill likelihood. According to DNV,
18 Psarros et al. (2010) point out that, although their findings coincide with previous studies (Norway 2008
and Thomas and Skjong 2009 as cited in Psarros et al. 2010), the results should be treated carefully.
33

LRFP data represent international tanker incident frequencies between 1990 and
2006 (Brandsæter and Hoffman 2010 p. 5-­‐49). A report entitled Study of Tug
Escorts in Puget Sound prepared by The Glosten Associates for the Washington
State Department of Ecology in 2004 discusses the current and historical use of
escort tugs at several international ports in a time period that overlaps with the
1990-­‐2006 data used by DNV in its estimate of tanker incidents for the ENGP. In
2004 when the study was released, escort tugs were required under state or
federal law in at least four locations in the United States and a further nine
locations voluntary used escort tugs for inbound and outbound tankers (Glosten
2004 p. 8-­‐9) (Table 21). DNV provides no evidence in the QRA that it adjusted the
LRFP dataset for ports where escort tugs are already in use. Therefore if these
adjustments were not made, the downward adjustments to LRFP data double
count mitigation impacts of tugs and underestimate the actual spill risk for ENGP
tankers.
Table 21: Escort Tug Usage at International Tanker Ports
Location Escort Tug Usage
Puget Sound, Washington Mandatory
Prince William Sound, Alaska Mandatory
San Francisco Bay, California Mandatory
Los Angeles/Long Beach, California Mandatory
Whiffenhead, Newfoundland Voluntary
Mongstad, Norway Voluntary
Rafsnes, Norway Voluntary
Brofjorden, Sweden Voluntary
Gothenburg, Sweden Voluntary
Porvoo, Finland Voluntary
Sullom Voe, Scotland Voluntary
Milford Haven, England Voluntary
Liverpool, England Voluntary
Source: Adapted from Glosten (2004 p. 8-­‐9)
4. Inadequate consideration of the corrosiveness of diluted bitumen associated with
internal corrosion that may increase spill frequency
The methodological approach estimating spill frequencies provides an inadequate
treatment of the corrosiveness of dilbit that contains higher sulphur
concentrations (EC OPD 2012). Enbridge does not incorporate the potential for
increased corrosion associated with transporting dilbit into return periods for
pipeline spills in Volume 7B nor does DNV include it in their methodology for
determining tanker and terminal spill return periods. The Bercha Group (2012b)
study evaluating pipeline failure frequencies that affect public safety recognizes
internal and external corrosion as potential causes for pipeline failure but
significantly reduces risk from corrosion by 80% without adequate justification.
Further, WorleyParsons (2012) states that the threat of internal and external
corrosion is negligible and thus excludes corrosion in their estimates of pipeline
spills.
34

Evidence suggests that sulphur concentrations in dilbit may increase the
corrosiveness of metals used for ENGP tanker, terminal, and pipeline operations.
Higher sulphur concentrations can present a corrosive threat to the integrity of
metals (Lyons and Plisga 2005) including metals used for pipelines, storage tanks,
and cargo holds. According to Environment Canada’s Oil Properties Database, the
sulphur content of West Texas Intermediate oil ranges between 0.34% to 0.57%
whereas the sulphur content for Athabasca Bitumen ranges from 4.41% to 5.44%
(EC OPD 2012), or between nine and 12 times higher than the benchmark of West
Texas Intermediate. Hence, the higher sulphur content in dilbit presents a
potential increased risk of failure from internal corrosion for metals used in the
construction of pipelines, tankers, and storage tanks for the ENGP.
Higher internal corrosion rates associated with the transportation of dilbit are
evident in the comparison of corrosion incidents between Alberta and US pipeline
systems. Between 1990 and 2005, internal corrosion accounted for nearly a
quarter (24.8%) of crude oil pipeline incidents in Alberta (AEUB 2007). In
comparison, internal corrosion incidents for crude oil pipelines in the US
represented less than one-­‐sixth (13.7%) of all pipeline incidents between 1986 and
2009 (US PHMSA 2012) 19 . Although these findings do not prove that increased
corrosiveness associated with transporting dilbit caused a higher rate of failure in
the Alberta pipeline system compared to the US system, the results illustrate that
more information is needed on the corrosiveness of dilbit to metals used for ENGP
operations.
In summary, the risk analysis should incorporate any potential increases in
corrosiveness from transporting and handling dilbit in an appropriate manner to
ensure that the potential for higher failure rates are quantified in probabilistic
terms. Increased corrosion could negate any risk reducing effects from mitigation
measures that purportedly decrease the likelihood of tanker, terminal, and pipeline
spills.
5. Failure to consider different vessel characteristics in tanker incident frequencies
There are two principle concerns related to vessel characteristics in the estimation
of tanker incident frequencies. First, DNV’s methodology in its marine shipping
QRA fails to differentiate between VLCC, Suezmax, and Aframax tankers for tanker
and terminal spills, even though Enbridge recognizes distinct maneuverability
differences among tanker classes. According to DNV, “The incident frequencies
derived from the LRFP data are considered to be valid for all three tanker classes
forecast to call at Kitimat Terminal. Tanker incident frequencies are influenced
more by the specific shipping route, than the type of tanker” (Brandsæter and
19 We calculate the proportion of internal corrosion incidents based on separate incident data from the
PHMSA for 1986-­‐2001 and 2002-­‐2009. Between 1986 and 2001, the PHMSA reported 1,263 total incidents,
168 (13.3%) of which were due to internal corrosion. Between 2002 and 2009, PHMSA reported 182 internal
corrosion incidents out of 1,291 total incidents or 14.1%. Both data sets represent internal corrosion
incidents for onshore crude oil pipelines.
35

Hoffman 2010 p. 5-­‐49). However, a report prepared for Enbridge by FORCE
Technology entitled Maneuvering Study of Escorted Tankers to and from Kitimat:
Real-­‐time Simulations of Escorted Tankers for a Terminal at Kitimat clearly
determines distinct differences among VLCC, Suezmax, and Aframax tankers for
several maneuverability characteristics including steering ability, turning ability,
and stopping ability. In terms of steering ability, FORCE determines that VLCCs
require over one and a half times the distance to conduct a zigzag test 20 than
Aframax tankers with both vessels in laden condition (see Figure 4-­‐1 in FORCE
2010). Similarly, for the turning ability test 21 for laden tankers, results show a
considerably larger area required to turn VLCCs than both Suezmax and Aframax
tankers when at full sea speed and at 10 knots (See Figures 4-­‐2 and 4-­‐2 in FORCE
2010). Finally, in terms of stopping ability, loaded VLCCs require nearly one and a
half times the stop distance compared to Aframax tankers and nearly twice the
distance than Suezmax tankers at 10 knots when the engines are running astern at
full power (Figure 4-­‐5 in FORCE 2010).
Based on the detailed findings in the report submitted by FORCE, there are indeed
very different maneuverability characteristics among VLCCs, Suezmax, and
Aframax tankers navigating shipping routes. Thus, DNV’s assumption of
uniformity among the three tanker classes fails to incorporate any of the distinct
handling capabilities of each tanker that could affect incident rates. Depending on
the nature of the LRFP data, the uniformity assumption potentially results in an
underestimate of particular incident occurrences for VLCCs compared to Suezmax
and Aframax tankers if VLCCs have a higher incident frequency and the LRFP
incident data largely consist of incident frequencies for Suezmax and Aframax
tankers. DNV’s failure to include proprietary LRFP data prevents any verification
of incident occurrence rates for the various tanker classes, however Eliopoulou
and Papanikolaou (2007) found that tanker incidents with serious consequences
or total losses and spillage rates between 1990 and 2003 were highest for VLCCs
compared to Aframax and Suezmax tankers.
A second consideration is the relative incident frequencies among different age
classes of tankers. A recent study from Eliopoulou et al. (2011) examines the
relationship between tanker age and accidents in tanker casualty data from the
LRFP database after the Oil Pollution Act of 1990. The authors determine that
incident rates for non-­‐accidental structural failure, or what DNV refers to as
foundering, vary significantly depending on the age of the double-­‐hull tanker.
Indeed, non-­‐accidental structural failure tanker incidents for double-­‐hull tankers
20 According to FORCE, tankers perform the zigzag test by “…commanding the rudder 10 deg. to port. When
the ship has changed its course 10 deg. from its initial course, the rudder is commanded 10 deg. to starboard.
When the ship has changed its course 10 deg. to starboard from its original course, the rudder is again
commanded 10 deg. to port.” (FORCE 2010 p. 19). The test completes after two zigzags.
21 Tankers perform the turning ability test by “…commanding the rudder 35 deg. to starboard while the
engine is running at the maximum load. As the ship starts to turn, a great part of the ship’s longitudinal speed
is transferred into a transverse (drift) speed that due to the great resistance reduces the longitudinal speed
significantly.” (FORCE 2010 p. 20).
36

anging between 16 and 20 years 22 are over 2.5 times higher compared to tankers
aged 11 to 15 years and over 4 times higher compared to tankers aged 6 to 10
years. In 2009, the authors estimate that the average age of double hull tankers in
the worldwide operational fleet was between 4 and 8 years. Papanikolaou et al.
(2009) estimate that, due to the young age of the worldwide tanker fleet, non-­‐
accidental structural failures could become significant after 2020, which is the date
that ENGP plans to be in full operation. Thus, historical incident frequency data
from the LRFP database may underestimate future incident rates for non-­‐
accidental structural failures for ENGP tankers since the methodological approach
in the QRA does not consider the likely increase in double-­‐hull tanker age during
the operating period of the ENGP.
In summary, DNV’s methodology should differentiate tanker incidents and spill
rates among the different tanker classes and different tanker ages. Failure to
consider different vessel characteristics results in a potential underestimate of
future tanker accidents.
Evaluation: There are five major deficiencies related to the reasonableness criterion and
thus this criterion is not met.
Reliability
Reliability: Appropriate analytical methods explicitly describe and evaluate sources of
uncertainty and variability that affect risk, and estimate the magnitudes of uncertainties
and their effects on estimates of risk.
The methodological approach estimating spill return periods in the ENGP regulatory
application contains two major deficiencies related to the reliability criterion:
1. Lack of confidence intervals that communicate uncertainty and variability in spill
estimates for tanker, terminal, and pipeline operations
The methodologies estimating tanker, terminal, and pipeline spills for the ENGP
fail to provide confidence intervals that characterize and communicate uncertainty
and variability. There are no confidence intervals for any of the spill estimates or
failure frequencies in Volume 7B, Volume 8C, the TERMPOL Study, the marine
shipping QRA prepared by DNV, the Bercha Group (2012a; 2012b; 2012c) studies,
or the WorleyParsons (2012) report. Since confidence intervals provide a measure
of the precision of a calculated value and describe the uncertainty surrounding an
estimate, the absence of confidence understates the degree of risk. Failure to
present confidence intervals implies that there is little or no uncertainty in spill
estimates and provides a false sense of confidence in the results. Thus spill
estimates for the ENGP fail to provide a measure of precision, fail to communicate
uncertainty in spill-­‐related data, and potentially mislead decision-­‐makers to
presume that there is certainty in the estimates even though this is not the case.
22 Enbridge has stated that it will not use tankers over 20 years of age in an effort to prevent hydrocarbon
spills (Enbridge 2010c).
37

2. Failure to complete a sensitivity analysis that effectively evaluates uncertainty
associated with spill estimates
The various technical reports addressing spill likelihood submitted for the ENGP
exclude comprehensive sensitivity analyses that measure the uncertainty of spill
estimates in a meaningful way. The analysis of pipeline spills in Volume 7B of the
ENGP application, the analyses prepared by the Bercha Group (2012a; 2012b;
2012c), and WorleyParsons (2012) do not contain sensitivity analyses.
DNV completed a sensitivity analysis in the marine shipping QRA although it
contains several deficiencies. First, the limited sensitivity analysis for tanker
traffic in the marine shipping QRA examines only four parameters: increased
scaling factor for grounding, increased traffic density, increased/decreased
number of tankers calling Kitimat Terminal, and extending segments 5 and 8 by
200 nm. DNV fails to examine sufficient increases in parameters, particularly
increases in tanker traffic associated with its expansion scenario that increases oil
pipeline throughput to 850 kbpd and condensate pipeline throughput to 275 kbpd
(Enbridge 2010b Information Request No. 3). This increased volume would result
in an increase in tanker traffic of 50%, much higher than the 14% increase in
tanker traffic used by DNV in its sensitivity analysis. The higher tanker traffic
would in turn significantly increase spill likelihood. Second, DNV does not examine
changes in critical parameters such as conditional probabilities and incident
frequencies to determine their effect on spill likelihood. Since incident frequencies
and conditional probabilities are multiplied together, any uncertainty in these
estimates propagate through to the final estimate of spill likelihood and could
result in significant changes to spill return periods. The uncertainty of these
critical parameters must be tested with a comprehensive sensitivity analysis.
Third, there is no sensitivity analysis for mitigated tanker operations, as well as
unmitigated and mitigated spills at the marine terminal. Fourth, the sensitivity
analysis for tanker spills only tests one parameter change at a time and does not
assess the impact of simultaneously changing multiple parameters. Fifth, other
important regulatory documents such as TERMPOL STUDY NO. 3.15: General Risk
Analysis and Intended Methods of Reducing Risk and Volume 8C: Risk Assessment and
Management of Spills – Marine Transportation do not refer to the limited sensitivity
analysis completed in the marine shipping QRA. Thus, the sensitivity analysis for
ENGP spill estimates evaluates only four parameters for the entire project,
ineffectively measures the magnitude of uncertainties, and results fail to appear in
main studies submitted by Enbridge for the ENGP regulatory application.
Evaluation: There are two major deficiencies related to the reliability criterion and thus
this criterion is not met.
Validity
Validity: Independent third-­‐party experts review and validate findings of the risk analysis
to ensure credibility, quality, and integrity of the analysis.
38

The absence of expert review for the majority of the ENGP regulatory application
related to spill return periods represents one major deficiency. Furthermore, the lack
of third party, independent expert reviewers for components of the risk analyses to
undergo expert review represents another major deficiency related to the validity
criterion.
1. Inadequate review and validation of spill estimates for tanker, terminal, and pipeline
operations
There is no evidence to suggest that the majority of findings in Volume 7B, Volume
8C, DNV’s marine shipping QRA, TERMPOL Study No. 3.15, the Bercha Group
(2012a; 2012b; 2012c) studies or the WorleyParsons study were reviewed or
validated by experts. Volume 7B and Volume 8C of the ENGP regulatory application
and the Bercha Group (2012c) study contain no reference to any expert peer
review or validation of estimates for tanker, terminal, and pipeline spills. DNV’s
methodological approach in the marine shipping QRA (and partly summarized in
the TERMPOL Study) underwent expert review to identify and evaluate hazards
that influence tanker risk, as well as validate local scaling factors that adjust tanker
operations to the BC study area. However, these two components represent only a
portion of the larger methodology estimating return periods for tanker spills and
DNV did not undertake expert review and validation of important findings related
to incident frequencies, conditional probabilities, and mitigation measures, as well
as return periods for tanker and terminal spills. The Bercha Group (2012a; 2012b)
claims that certain components of its methodological approach determining the
failure frequency for spills at Kitimat Terminal and pipeline ruptures were
discussed with experts. Yet, the Bercha Group provides no evidence of any
consultations with experts in either report and thus it remains unclear how expert
judgment was used in the reports and whether or not the experts were
independent. With regards to the WorleyParsons report, sections of the pipeline
failure frequency assessment underwent expert judgment, particularly failure
frequencies for incorrect operations and geohazards, although these frequencies
represent only two of the eight hazards for which spill frequencies were developed
and there are issues related to the independence of the experts.
The absence of review and validation of findings for the various methodological
approaches raises concerns over the credibility, quality, and integrity of spill
estimates in the ENGP regulatory application. Validation of results ensures that all
critical assumptions and uncertainties are acknowledged and documented,
appropriate models, methods, and data are used, and analysis is reproducible by
independent parties (IALA 2008).
2. Lack of third party, independent experts in the review and validation of spill estimates
A second major deficiency concerns the lack of third party, independent review
and validation of the findings that underwent expert review. The two studies to
undergo review, i.e. DNV’s marine shipping QRA and the WorleyParsons semi-­‐
39

quantitative risk analysis, were evaluated by experts affiliated with both
organizations. According to IALA (2008), third party reviewers such as
universities and government agencies should assess technical documentation to
confirm the integrity of the analysis and contribute credibility to the results.
In their attempt to identify hazards that affect tanker incident frequencies, DNV
consulted with local experts in marine transportation on the BC coast (Brandsæter
and Hoffman 2010 p. 4-­‐40) and held meetings and interviews with some local
stakeholders (Brandsæter and Hoffman 2010 p. 4-­‐46). Once local stakeholders
identified hazards in the workshop, experts evaluated hazards of the proposed
tanker routes that would be used by ENGP. However, the four experts evaluating
hazards consisted of three DNV employees and one employee of WorleyParsons, a
consultant hired by Enbridge (Brandsæter and Hoffman 2010 p. 4-­‐44). DNV then
incorporated the findings of the hazard identification process into tanker incident
frequencies by scaling world average frequencies in order to represent the BC
study area. DNV held another workshop where it conducted a peer review to
validate the findings of the process to determine appropriate local scaling factors
but the review and validation of local scaling factors lacked independence. Two of
the five experts involved in the review authored the marine shipping QRA
submitted by DNV (Brandsæter and Hoffman 2010 p. 5-­‐52), while one reviewer
was at one time a Principal Consultant with DNV Maritime Solutions (DNV 2008),
and another reviewer is affiliated with DNV (DNV 2004).
The WorleyParsons report examining risks of a pipeline spill also uses expert
review that lacks independence. WorleyParsons contracted AMEC to determine
the failure frequency of geohazards such as rock falls, stream flow and erosion, and
avalanches, and AMEC uses expert judgment to assess the likelihood of a pipeline
spill from geohazards (AMEC 2012 p. 2). Three of the four expert panel members
are affiliated with the organizations that either authored the report or directly
provided analysis for the report (AMEC 2012 p. 6). Similarly, Dynamic Risk
Assessment Systems, which completed analysis for the WorleyPasons study,
conducted a workshop to determine data requirements for estimating spill
likelihood. Every attendee at the workshop was either a representative of
Enbridge, AMEC, or WorleyParsons (DRAS 2012 p. 2).
In summary, the expert review and validation of findings undertaken by DNV and
WorleyParsons lack independence. We do not question either the expertise or
integrity of individuals that participated in the review, we simply emphasize that
there is a lack of third party, independent review and validation of spill estimates
for the ENGP.
Evaluation: There are two major deficiencies related to the validity criterion and thus this
criterion is only not met.
40

Risk Acceptability
Risk Acceptability: Stakeholders identify and agree on acceptable levels of risk in a
participatory, open decision-­‐making process and assess risk relative to alternative means
of meeting project objectives.
Methods estimating spill return periods in the ENGP regulatory application contain
three major deficiencies related to the risk acceptability criterion:
1. Failure to define risk acceptability in terms of the needs, issues, and concerns of
stakeholders potentially impacted by the project
The methodological approaches in the ENGP regulatory application all define risk
in technical terms. Risk acceptability in the marine shipping QRA has an implicit
technical definition that compares spill forecasts for the ENGP derived from
historical international data with spill frequencies representative of the ‘world
average’. DNV compares the unmitigated return period of 79 years for
oil/condensate spills to the international spill return period of 74 years and claims
that risk “is slightly better than the world average” and that the risk “...may be
acceptable compared to existing international operations…” (Brandsæter and
Hoffman 2010 p. 7-­‐116) 23 . Risk analyses completed by the Bercha Group (2012a;
2012b; 2012c) and WorleyParsons also fail to adequately assess and define risk
acceptability in terms of the needs, issues, and concerns of stakeholders potentially
impacted by the project. The Bercha Group studies all define risk acceptability
based in part on an algorithm that includes several variables such as the
probability of an individual experiencing fatality and the probability of an oil spill
from an accident, among others. The studies then compare the results of the
algorithm to pre-­‐determined levels of risk acceptability to determine whether or
not the risk is significant. Similarly, the WorleyParsons report determines the
significance of a threat or hazard in terms of data availability and quality, whereby
threats or hazards that cause pipeline failure are assessed as significant or
insignificant depending on historical data. Finally, Enbridge does not discuss nor
define acceptable levels of risk in Volume 7B for the likelihood of pipeline spills,
which implies that all levels of risk from pipeline spills are acceptable. A technical
definition of risk limits risk analysis to data, methods, and assumptions of the
analysts preparing the assessment and fails to incorporate the goals, objectives,
and concerns of stakeholders potentially impacted by risk.
2. Failure to effectively consult with stakeholders to determine a level of acceptable risk
for tanker, terminal, and pipeline spills
Technical information cannot provide the basis for determining acceptable levels
of risk (BC MoELP 2000). Indeed, risk acceptability must materialize from an open
dialogue with stakeholders that builds trust and provides confidence in the results
(IALA 2008). Determining acceptable levels of risk should be an iterative process
23 The unmitigated return period of 79 years referenced by DNV on p. 7-­‐116 of Brandsæter and Hoffman
(2010) is inconsistent with the unmitigated return period of 78 years referenced on pages 7-­‐107, 1-­‐108, and
page 137 of Brandsæter and Hoffman (2010).
41

that begins with stakeholder consultations early in the process and continues
throughout the process to address any residual risk after appropriate mitigation
measures have been proposed (IALA 2008). If stakeholders cannot agree on
acceptable levels of risk, the proposed activity may need to be abandoned (IALA
2008). Furthermore, the magnitude of adverse impacts is an important
consideration that affects risk acceptability. Acceptable risk for a major oil spill
will likely be lower than the risk for a smaller spill that causes less significant
impacts and may be lower than the risk accepted in other jurisdictions if the
attitudes of those impacted are more risk averse.
Stakeholders have not collectively defined and agreed on acceptable levels of risk
associated with ENGP tanker, terminal, and pipeline operations. In the marine
shipping QRA, DNV consults local experts familiar with marine transportation on
the BC coast to identify the influence of local hazards on spill frequencies
(Brandsæter and Hoffman 2010 p. 4-­‐40) and holds local meetings and interviews
with some local stakeholders to discuss shipping routes, local hazards, and
conditions that should be included in the analysis (Brandsæter and Hoffman 2010
p. 4-­‐46). Yet, identifying hazards and their potential risks is not equivalent to
determining an acceptable level of risk agreed on by all stakeholders potentially
impacted by a spill. Similarly, there is no discussion of any stakeholder
consultations in Volume 7B, the Bercha Group (2012a; 2012b; 2012c) studies, or
the WorleyParsons report, suggesting that stakeholder concerns are not included
in defining risk acceptability.
3. Inadequate assessment and comparison of risks associated with project alternatives
A third major deficiency concerns Enbridge’s inadequate approach to identifying
and comparing alternative approaches to meeting the objective of the ENGP, which
is “…to provide access for Canadian oil to large and growing international markets,
comprising existing and future refiners in Asia and the United States West Coast”
(Enbridge 2010b Vol. 1 p. 1-­‐3). In its evaluation of alternatives, Enbridge fails to
justify the ENGP in its proposed configuration, fails to adequately consider risks
associated with alternative project configurations, and does not examine and
compare risks associated with possible project alternatives beyond simply
reconfiguring the ENGP.
In Volume 1 of its regulatory application, Enbridge outlines several alternative
locational configurations of the ENGP prior to selecting the ENGP in its current
configuration but provides insufficient information supporting the preferred
option. According to Enbridge (2010b Vol. 1 p. 4-­‐3), it considered different
locations near Edmonton and Fort McMurray for the inland pipeline terminus in
Alberta and examined marine terminals in numerous locations, eventually
narrowing candidate sites to Prince Rupert and Kitimat. Enbridge determined that
a pipeline terminus near Edmonton and a marine terminal in Kitimat were the
preferred alternatives based on evaluative criteria that includes project lifecycle
costs, acceptability to shippers, suitability for tankage, safety of tanker operations,
constructability of project infrastructure, construction and operation safety, and
42

the likelihood that environmental and socioeconomic interests are affected
(Enbridge 2010b Vol. 1. p. 4-­‐1). However, Enbridge fails to provide adequate
justification that each alternative was evaluated according to its criteria and omits
any detailed comparison of project alternatives required to identify the preferred
alternative.
The assessment of alternatives also fails to explicitly evaluate and consider spill
probabilities for tanker, terminal, and pipeline spills for each alternative project
configuration along with the magnitude of corresponding impacts should a spill
occur. Enbridge consulted analysis completed by Fisheries and Oceans Canada in
the 1970s that examines the potential effects of oil spills at 11 west coast ports
prior to narrowing down marine terminal options to Kitimat and Prince Rupert
(Enbridge 2010b Vol. 1 p. 4-­‐3). Nevertheless, Enbridge’s assessment represents an
inadequate comparison of risks associated with each alternative terminal site that
fails to assess and compare the likelihood of tanker and pipeline spills in a
comprehensive manner.
On a broader scale, Enbridge does not identify nor compare viable alternatives that
meet the stated purpose of the project beyond simply reconfiguring component
parts of the ENGP. Realistic project alternatives should be identified, evaluated,
and compared and the analysis should include a risk profile for each alternative
that enables a comparison of the risks of each candidate project. There are viable
alternatives to shipping crude oil from the Western Canada Sedimentary Basin to
market that involve no risk from oil tanker spills (Gunton and Broadbent 2012a)
and thus determining risk acceptability for tanker spills can be eliminated. A
comprehensive comparison of the various project alternatives ensures that
decision-­‐makers have adequate information to determine the project that satisfies
the public interest of all Canadians while minimizing environmental, cultural,
social, and economic risks. Enbridge has not completed such a comparison.
Evaluation: There are three major deficiencies related to the criterion for risk
acceptability and thus this criterion is not met.
4.1. Summary of Qualitative Assessment of Spill Estimates for the ENGP
Table 22 summarizes the results of our evaluation of the methodological approach used
to determine spill return periods for the ENGP. Our analysis determined that none of
the seven best practice criteria have been met in the ENGP regulatory application. The
analysis also revealed a total of 28 major deficiencies in the approaches estimating
ENGP tanker, terminal and pipeline spills. Based on our analysis, the information
presented by Enbridge in the regulatory application for the ENGP does not accurately
and effectively communicate the likelihood of adverse environmental effects resulting
from a spill required under CEAA.
43

Table 22: Results for the Qualitative Assessment of Spill Estimates for the ENGP
Criterion Major Deficiencies Result
Transparency
Documentation fully and
effectively discloses supporting
evidence, assumptions, data
gaps and limitations, as well
as uncertainty in data and
assumptions, and their
resulting potential
implications to risk
Replicability
Documentation provides
sufficient information to allow
individuals other than those
who did the original analysis
to obtain similar results
Clarity
Risk estimates are easy to
understand and effectively
communicate the nature and
magnitude of the risk in a
manner that is complete,
informative, and useful in
decision making
Reasonableness
The analytical approach
ensures quality, integrity, and
objectivity, and meets high
scientific standards in terms of
analytical methods, data,
assumptions, logic, and
judgment
Reliability
Appropriate analytical
methods explicitly describe
and evaluate sources of
uncertainty and variability
that affect risk, and estimate
the magnitudes of
uncertainties and their effects
on estimates of risk
Validity
Independent third-­‐party
experts review and validate
findings of the risk analysis to
ensure credibility, quality, and
integrity of the analysis
Risk Acceptability
Stakeholders identify and
agree on acceptable levels of
risk in a participatory, open
decision-­‐making process and
assess risk relative to
alternative means of meeting
project objectives
1. Lack of supporting evidence for LRFP tanker incident frequency data
2. Insufficient evidence supporting conditional probabilities for tanker spills
3. Limited transparency in the application of local scaling factors to the BC
study area
4. Lack of information comparing incident frequencies at Kitimat Terminal to
marine terminals in Norway
5. Lack of transparency in documenting how mitigation measures will reduce
the likelihood of tanker and terminal spills
6. Inadequate evidence supporting the reduction of spill frequencies for
pipeline operations
Insufficient proprietary data and information required to replicate:
7. Incident frequencies for tanker accidents
8. Scaling factors for ENGP routes
9. Probabilities and spill volume estimates for tanker spills,
10. Mitigation measures for tanker and terminal operations
11. Pipeline incident frequencies
12. Failure to effectively communicate the probability of spills over the life of
the project
13. Failure to combine estimates for tanker, terminal, and pipeline spills to
present a single spill estimate for the entire project
14. Failure to determine the likelihood of all spill size categories that can have
a significant adverse environmental impact
15. Lack of clear communication of spill estimates generated in detailed
technical data reports in the main ENGP application
16. Failure to address the effective implementation of mitigation measures
that reduce risk
17. Inappropriate definition of the study area to estimate spill return periods
for tanker operations
18. Reliance on tanker incident frequency data that underreport incidents by
between 38 and 96%.
19. Potential double-­‐counting of mitigation measures for LRFP tanker incident
data
20. Inadequate consideration of the corrosiveness of diluted bitumen
associated with internal corrosion that may increase spill probability
21. Failure to consider different vessel characteristics in tanker incident
frequencies
22. Failure to provide confidence intervals that communicate uncertainty and
variability in spill estimates for tanker, terminal, and pipeline operations
23. Failure to complete an adequate sensitivity analysis that effectively
evaluates uncertainty associated with spill estimates
24. Inadequate expert review and validation of spill estimates for tanker,
terminal, and pipeline operations
25. Experts in the review and validation of spill estimates lack independence
and third-­‐party status
26. Failure to define risk acceptability in terms of the needs, issues, and
concerns of stakeholder potentially impacted by the project
27. Failure to effectively consult with stakeholders to determine a level of
acceptable risk for tanker, terminal, and pipeline spills
28. Inadequate assessment and comparison of risks associated with project
alternatives
Not
Met
Not
Met
Not
Met
Not
Met
Not
Met
Not
Met
Not
Met
44

5. Extended Sensitivity Analysis for ENGP Spill Estimates
In this section we undertake a sensitivity analysis to assess the impact of the
weaknesses in the methodological approach used to estimate return periods for the
ENGP. We caution that our sensitivity analysis does not attempt to addresses all the
weaknesses identified in the qualitative assessment. Instead, we demonstrate the
impact of different assumptions on spill return periods for ENGP tanker, terminal, and
pipeline operations.
5.1. Sensitivity Analysis for Spill Return Periods for Tanker Operations
Our extended sensitivity analysis for ENGP tanker operations examines effects of the
following input parameters on unmitigated and mitigated return periods for an oil or
condensate spill:
• Testing the combined impact of the four sensitivity parameters identified by
DNV in its sensitivity analysis
• Using incident frequencies derived from LRFP data without scaling for BC
conditions, thus assuming that BC marine spill risk factors for ENGP tanker
traffic conform to the ‘world average’
• Increasing conditional probabilities by five percentage points
• Increasing tanker traffic to transport pipeline throughput of 1,125 kbpd (850
kbpd of oil and 275 kbpd of condensate)
• Increasing tanker traffic to represent replacing VLCCs with smaller Suezmax
and Aframax tankers to transport throughput of 1,125 kbpd of oil and
condensate
• Testing the combined effect of the parameters used in our extended sensitivity
analysis.
We follow a similar methodological approach as DNV in order to compare the results of
our sensitivity analysis with the sensitivity analysis for tanker spill return periods in
the marine shipping QRA. However, one area where our analysis differs is the
treatment of tanker traffic. In its sensitivity analysis, DNV compares the relative
difference between routes by assuming that 220 tankers travel each route every year.
Our sensitivity analysis uses the forecast traffic distribution identified by DNV for
tanker traffic along each route. Thus, return periods in our extended sensitivity
analysis compare to the unmitigated and mitigated spill return periods of 78 years and
250 years determined by DNV in the marine shipping QRA that reflect forecast traffic
on each route. We have done this to simplify the presentation of findings by showing
them for one route as opposed to three routes. Another area in which our sensitivity
analysis differs from that of DNV is the direction of changes to input parameters. Our
analysis only examines changes to input parameters that reduce return periods,
thereby increasing the likelihood of a spill. We recognize that additional sensitivity
analysis should be completed on the range of input parameters to determine their
effect on spill return periods.
45

Sensitivity 1: Combined Parameters from the DNV Sensitivity Analysis
The first sensitivity analysis combines the four sensitivity parameters identified by
DNV in its sensitivity analysis. We examine the collective effect of an increase to local
scaling factors for powered and drift grounding by 20%, an increase in scaling factors
ranging from 0% to 50% for traffic that affects collision frequency, an increase in the
number of tankers calling at Kitimat to 250 per year, and an extension of segments 5
and 8 to 200 nm. Combining sensitivity parameters significantly reduces return
periods to 56 years and 171 years for unmitigated and mitigated spills, respectively
(Table 23). Our analysis does not consider potential synergies among the four
parameters that may further decrease spill return periods.
Table 23: Sensitivity Analysis Combining Sensitivity Parameters Identified by DNV
Sensitivity Parameter
Return Period (in years)
Unmitigated Mitigated
Baseline 78 250
Combined Sensitivities 56 171
Source: Based on calculations from Brandsæter and Hoffman (2010).
Sensitivity 2: Unscaled Incident Frequencies
A second sensitivity analysis examines the effect of local scaling factors on unmitigated
and mitigated spill return periods for oil or condensate spills by using unscaled
incident frequencies. We assume all scaling factors for each incident type are set to 1.0,
which implies that tanker operations associated with the ENGP are consistent with the
‘world average’ that DNV claims is representative of LRFP data (Brandsæter and
Hoffman 2010 p. 5-­‐51). Based on this assumption, return periods decrease
significantly compared to the baseline, particularly mitigated return periods that
decrease over 150 years from 250 years to 83 years under ‘world average’ conditions
for (Table 24).
Table 24: Sensitivity Analysis based on Unscaled Incident Frequencies
Sensitivity Parameter
Return Period (in years)
Unmitigated Mitigated
Baseline 78 250
Unscaled Frequencies 54 83
Source: Based on calculations from Brandsæter and Hoffman (2010).
Sensitivity 3: Increased Conditional Probabilities
We examine the effect of changes in conditional probabilities on spill return periods.
In the consequences assessment, DNV fails to provide evidence to support the values
assigned for each incident type. To test the sensitivity of conditional probabilities on
return periods, we increase conditional spill probabilities by five percentage points for
grounding, collision, and fire/explosion incidents (Table 25).
46

Table 25: Five-­‐Percentage Point Increase in Conditional Probabilities for Sensitivity Analysis
Incident Type
Baseline Conditional Probabilities
Laden In Ballast
Increased Conditional Probabilities
Laden In Ballast
Grounding 32.7% 6.4% 37.7% 11.4%
Collision 19.1% 2.6% 24.1% 7.6%
Foundering* 100% 100% n/a n/a
Fire/Explosion 27.0% 7.6% 32.0% 12.6%
Source: Based on calculations from Brandsæter and Hoffman (2010).
* DNV assumes that foundering incidents result in a 100% loss and thus conditional probabilities cannot increase further.
We estimate spill return periods by multiplying increased conditional probabilities by
scaled incident frequencies, tanker traffic distribution and distances for each route. As
shown in Table 26, the return period for an unmitigated oil or condensate spill decreases
as much as 16 years from 78 years to 62 years, while the return period for a mitigated
oil/condensate spill decreases 54 years compared to the baseline.
Table 26: Sensitivity Analysis for Increased Conditional Probabilities by Five Percentage Points
Sensitivity Parameter
Return Period (in years)
Unmitigated Mitigated
Baseline 78 250
Increased Conditional Probabilities 62 196
Source: Based on calculations from Brandsæter and Hoffman (2010).
Sensitivity 4: Increased Tanker Traffic to Transport Pipeline Throughput of 1,125 kbpd
DNV examines an increase in tanker traffic to 250 tankers per year but does not
provide any rationale for restricting the sensitivity to this number. A more realistic
approach is to set traffic increases to the level required to transport increased oil and
condensate pipeline throughput equivalent to the expansion scenario identified by
Enbridge in its response to Information Request No. 3. In its response, Enbridge
outlines a four-­‐phase expansion of oil pipeline capacity to 850 kbpd and condensate
pipeline expansion to 275 kbpd (see Table 27) that includes additional pump stations
and infrastructure required to support increased throughput (Enbridge 2010b
Information Request No. 3 p. 5). Under the expansion scenario, total combined
throughput of both oil and condensate pipelines would total 1,125 kbpd, which would
require approximately 237 oil tankers and 94 condensate tankers. The increase in
tanker traffic from 220 to 331 tankers per year under Enbridge’s phased expansion
represents a 50% increase in tanker traffic.
Table 27: Potential Expansion Scenario of Oil and Condensate Pipelines
Pipeline
Throughput
Phase I (kbpd) Phase II (kbpd) Phase III (kbpd) Phase IV (kbpd)
Oil 525 600 750 850
Condensate 193 250 275 n/a
Source: Enbridge (2010b Information Request No. 3 p. 9-­‐10).
Table 28 presents unmitigated and mitigated spill return periods for tanker traffic
required to transport increased oil and condensate throughput of 1,125 kbpd. We
47

assume that although the number of tankers increases to 331 per year, the proportion
of tankers required to transport 1,125 kbpd remains the same for each route as the
distribution identified by DNV (See Appendix A for tanker traffic per route). Return
periods decrease 26 years for unmitigated oil/condensate spills and 81 years for
mitigated spills when tanker traffic increases to transport ENGP throughput of 1,125
kbpd.
Table 28: Sensitivity Analysis for Tankers Transporting 1,125 kbpd
Sensitivity Parameter
Return Period (in years)
Unmitigated Mitigated
Baseline 78 250
1,125 kbpd (331 Tankers)* 52 169
Source: Based on calculations from Brandsæter and Hoffman (2010).
* ENGP throughput of 1,125 kbpd requires 237 tankers to transport 850 kbpd and 94 tankers to transport 275 kbpd of condensate.
Sensitivity 5: Replacing VLCCs with Smaller Suezmax and Aframax Tankers to Transport
1,125 kbpd of Oil and Condensate
In the consequence assessment, DNV acknowledges increased risk associated with
using a larger number of lower capacity tankers to transport the same amount of cargo
(Brandsæter and Hoffman 2010). However, the sensitivity analysis in the marine
shipping QRA does not provide estimates of spill return periods associated with using
smaller Suezmax and Aframax tankers instead of VLCCs. According to our calculations,
the overall tanker fleet would increase 22% from 220 tankers (149 oil and 71
condensate) to 268 tankers per year (197 oil and 71 condensate) if smaller Suezmax
and Aframax tankers replace VLCCs. Since the previous sensitivity analysis for the
expansion scenario already tests the effects of an increase in tanker traffic, we combine
the foregone use of VLCCs with the expansion scenario and estimate the effect on spill
return periods.
The scenario in which Suezmax and Aframax tankers transport 1,125 kbpd requires a
total of 412 tankers per year to transport 850 kbpd of oil (318 tankers) and 275 kbpd
of condensate (94 tankers). We assume that the proportion of Suezmax, and Aframax
tankers remains the same as the distribution identified by DNV, although the number
of tankers increases in order to transport the same amount of oil and condensate in the
absence of VLCCs. A shown in Table 29, the increase in traffic associated with the
foregone use of VLCCs under the expansion scenario decreases the return period for an
unmitigated oil or condensate spill by 35 years and decreases the mitigated
oil/condensate spill return period by 108 years.
Table 29: Sensitivity Analysis for Smaller Tankers Transporting Throughput of 1,125 kbpd
Sensitivity Parameter
Return Period (in years)
Unmitigated Mitigated
Baseline 78 250
No VLCCs at 1,125 kbpd (412 Tankers)*
Source: Based on calculations from Brandsæter and Hoffman (2010).
43 142
* ENGP throughput of 1,125 kbpd without the use of VLCCs requires 318 tankers to transport 850 kbpd of oil and 94 tankers to
transport 275 kbpd of condensate.
48

Sensitivity 6: Aggregating Parameters of the Extended Sensitivity Analysis
Similar to combining sensitivity parameters for the sensitivity analysis completed by
DNV, we aggregate our sensitivity parameters into a single parameter to examine the
combined effects on return periods. We combine the following parameters of our
extended sensitivity analysis:
• Unscaled incident frequencies derived from LRFP data that assume
oil/condensate spills associated with ENGP tanker traffic conform to the ‘world
average’
• Increased tanker traffic to 412 tankers per year as a result of increased
throughput of 1,125 kbpd (850 kbpd of oil and 275 kbpd of condensate) and the
replacement of VLCCs with smaller Suezmax and Aframax tankers
• Increased conditional probabilities by five percentage points.
Combining these parameters decreases the return period for an unmitigated
oil/condensate spill 55 years to one spill every 23 years (Table 30). The return period
for a mitigated oil or condensate spill decreases significantly by 216 years from a
baseline of 250 years to 34 years when we combine the parameters of our extended
sensitivity analysis.
Table 30: Sensitivity Analysis Aggregating Parameters from the Extended Sensitivity Analysis
Sensitivity Parameter
Return Period (in years)
Unmitigated Mitigated
Baseline 78 250
Aggregate Parameters 23 34
Source: Based on calculations from Brandsæter and Hoffman (2010).
5.2. Sensitivity Analysis for Spill Return Periods for Operations at Kitimat Terminal
DNV did not complete a sensitivity analysis for oil or condensate spills at Kitimat
Terminal. To address this deficiency, we conduct a sensitivity analysis for marine
terminal operations based on a replication of spill return periods for oil and
condensate tankers accessing Kitimat Terminal. Our sensitivity analysis uses the
methodology and data presented by DNV in its marine shipping QRA. DNV presents
return periods for small and medium size oil or condensate spills and uses spill size
categories from the International Tanker Owners Pollution Federation where a small
spill is less than 10 m 3 and a medium spill ranges between 10 and 1,000 m 3 . Spill sizes
for this sensitivity analysis represent small or medium spills (i.e. spills less than 1,000
m 3 ), which DNV refers to as any size spills (Brandsæter and Hoffman 2010 p. 136).
The focus of our sensitivity analysis is on cargo transfer operations since, according to
DNV, loading and unloading operations present the greatest chance of a spill
(Brandsæter and Hoffman 2010). We examine changes in two key parameters
associated with cargo transfer operations.
49

Sensitivity 1: Increased Tanker Traffic to Transport Pipeline Throughput of 1,125 kbpd
We increase annual tanker traffic 50% from 220 tankers to 331 tankers required to
transport increased pipeline throughput of 1,125 kbpd (237 tankers required to
transport 850 kbpd of oil and 94 tankers required to handle 275 kbpd of condensate).
An increase to 331 tankers per year decreases the return period for an unmitigated
oil/condensate spill of any size by 10 years and decreases the mitigated return period
by 21 years (Table 31).
Table 31: Sensitivity Analysis for Increased Marine Terminal Throughput of 1,125 kbpd
Sensitivity Parameter
Return Period (in years)
Unmitigated Mitigated
Baseline Cargo Transfer Operations 29 62
1,125 kbpd (331 Tankers)* 19 41
Source: Based on calculations from Brandsæter and Hoffman (2010).
* ENGP throughput of 1,125 kbpd requires 237 tankers to transport 850 kbpd of oil and 94 tankers to transport 275 kbpd of
condensate.
Sensitivity 2: Replacing VLCCs with Smaller Suezmax and Aframax Tankers to Transport
1,125 kbpd of Oil and Condensate
For the second parameter, we calculate the effect of increased tanker traffic on spill
return periods at the marine terminal if smaller Suezmax and Aframax tankers replace
VLCCs in the transportation of 1,125 kbpd of oil and condensate. This scenario
requires 318 Suezmax and Aframax oil tankers to replace the fleet of 237 oil tankers
inclusive of VLCCs required to transport 850 kbpd of oil, while condensate tankers
remain constant at 94 tankers per year in order to transport 275 kbpd of condensate 24 .
Compared to baseline conditions of 220 tankers, the return period for an
oil/condensate spill decreases 14 years for an unmitigated spill and decreases 29 years
for a mitigated spill when 412 Suezmax and Aframax tankers transport 1,125 kbpd
(Table 32).
Table 32: Sensitivity Analysis for Smaller Tankers Transporting Throughput of 1,125 kbpd at the
Marine Terminal
Sensitivity Parameter
Return Period (in years)
Unmitigated Mitigated
Baseline Cargo Transfer Operations 29 62
No VLCCs at 1,125 kbpd (412 Tankers)* 15 33
Source: Based on calculations from Brandsæter and Hoffman (2010).
* ENGP throughput of 1,125 kbpd without the use of VLCCs requires 318 tankers to transport 850 kbpd of oil and 94 tankers to
transport 275 kbpd of condensate.
5.3. Sensitivity Analysis for Spill Return Periods for Pipeline Operations
Enbridge did not complete a sensitivity analysis for spill return periods for the crude
oil and condensate pipelines between Bruderheim, AB and Kitimat, BC. To examine the
24 According to Brandsæter and Hoffman (2010 p. 7-­‐105), no VLCCs are expected to transport condensate
and thus the foregone use of VLCCs does not affect condensate tanker traffic.
50

effects of different assumptions on spill likelihood, we use two different data sets to
estimate the frequency of oil and condensate pipeline spills for the ENGP:
• Spill data from the Enbridge liquids pipeline system from 2002 to 2010
• Spill data from the NEB for pipebody and operational leaks > 1.5 m 3 between
2000 and 2009.
We caution that the pipeline spill sensitivity is based on historical performance of
existing pipelines that may not be representative of the risks associated with the ENGP.
Sensitivity 1: Enbridge Pipeline Spill Data from 2002 to 2010
We estimate return periods for the ENGP using spill frequency data for the number of
spills per annual volume of liquid shipped on the Enbridge liquids pipeline system.
According to data submitted by Enbridge (2012a) in the regulatory review process for
the ENGP, a total of 608 reportable spills occurred on the Enbridge liquids pipeline
system between 2002 and 2010 releasing an average of 26 m 3 (nearly 162 bbl) per
spill (Table 33).
Table 33: Historical Spill Data for the Enbridge Pipeline System (2002 -­‐ 2010)
Year
Annual Volume
Shipped
(in barrels)
Number of
Spills
Spill
Volume (m 3)
Spills per Barrel
Shipped*
Liquid Released
per Spill* (m 3)
2002 733,440,396 46 2,334 6.27E-­‐08 51
2003 867,686,223 58 1,014 6.68E-­‐08 17
2004 1,123,230,798 64 495 5.70E-­‐08 8
2005 1,060,531,358 70 1,562 6.60E-­‐08 22
2006 1,198,602,588 62 864 5.17E-­‐08 14
2007 1,199,657,322 59 2,187 4.92E-­‐08 37
2008 1,256,268,675 80 426 6.37E-­‐08 5
2009 1,333,947,614 89 1,329 6.67E-­‐08 15
2010 1,487,709,454 80 5,425 5.38E-­‐08 68
Total 10,261,074,428 608 15,638 5.93E-­‐08 26
Source: Enbridge (2012a).
* Calculated from Enbridge (2012a).
Based on the spill frequency of 5.93E-­‐08 per barrel shipped and proposed shipments of
525 kbpd of crude oil and 193 kbpd of condensate for the ENGP, over 15 oil and
condensate spills could occur from the ENGP in any given year (Table 34). The over 15
spills could release an average of 399 m 3 (2,512 bbl) of oil and condensate per year
once the ENGP becomes operational. Thus Enbridge’s own data shows that the return
period for a crude oil or condensate spill is less than one year.
51

Table 34: Sensitivity Analysis for ENGP Pipelines Using Spill Data for the Enbridge Liquids
Pipeline System
Type of
Pipeline
Spill Frequency
(spills per bbl
shipped)
Number of
Spills for ENGP
(per year)
Spill Volume
(m 3)
Return Period
(in years)
Crude Oil
Condensate
5.93E-­‐08
11.4
4.2
292
107
0.1
0.2
Crude Oil and Condensate Pipelines
Source: Calculated from Enbridge (2012a).
15.5 399 0.1
Note: We calculate the number of spills based on crude and condensate deliveries of 525 kbpd and 193 kbpd, respectively.
Sensitivity 2: Unmitigated NEB Pipeline Leak Data from 2000 to 2009
We estimate spill frequency for the ENGP based on data from the NEB pipeline system
between 2000 and 2009, which represents pipebody and operational leaks. A
pipebody leak is any leak from the body of the pipe and includes cracks and pinholes
(NEB 2011 p. 13) and operational leaks are leaks from pipeline components such as
valves, pumps, and storage tanks (NEB 2011 p. 15). The NEB uses a spill size
classification of > 1.5 m 3 (9 bbl) for pipebody and operational leaks. Table 35 contains a
history of pipebody liquid leaks from 2000 to 2009, which is the only available data
from the NEB 25 .
Table 35: Historical NEB Pipebody and Operational Leak Data from 2000 to 2009
Year
Length of NEB
Liquid Pipeline (km)
Liquid
Pipebody
Leaks > 1.5 m 3
Liquid
Operational
Leaks > 1.5 m 3
Total
Leaks
> 1.5 m 3
Leak Frequency
> 1.5 m 3 *
(spills per km-­‐year)
2000 13,219 0 2 2 0.00015
2001 16,165 2 4 6 0.00037
2002 14,803 2 9 11 0.00074
2003 15,245 0 1 1 0.00007
2004 15,012 0 5 5 0.00033
2005 14,269 2 3 5 0.00035
2006 15,566 4 7 11 0.00071
2007 14,368 4 4 8 0.00056
2008 15,715 0 6 6 0.00038
2009 14,442 2 4 6 0.00042
Total 148,804 16 45 61 0.00041
Source: NEB (2011).
* Calculated from NEB (2011).
Based on NEB pipeline leak data and the length of the ENGP oil and condensate
pipelines, the annual spill frequency for spills > 1.5 m 3 (9 bbl) is 0.5 spills per pipeline
(Table 36), which represents a spill return period of two years for each of the oil and
condensate pipelines. If spill frequencies for crude oil and condensate pipelines are
25 In response to our request for historical annual data on the number of pipe body liquid releases > 1.5 m 3
and the total volume of leaks per year, the NEB stated “…unfortunately we do not have historical (pre-­‐2000)
data analyzed in this way. To perform the requested analysis of data older than 2000 would require weeks of
work, as we would need to physically read and analyze each individual investigation report in order to
determine if it met the requested criteria and then compile the data from there.” (Caza 2012).
52

combined, the NEB data suggests that a spill > 1.5 m 3 could occur every year the
Northern Gateway Project is in operation.
Table 36: Sensitivity Analysis for ENGP Pipelines Using NEB Data for Leaks >1.5m 3
Operating
Spill Frequency Number of Leaks >1.5 m
Period
(spills per km-­‐year)
3 Return Period
for ENGP (per year)
(in years)
Crude Oil
Condensate
0.00041
0.5
0.5
2
2
Crude Oil and Condensate Pipelines
Source: Calculated from NEB (2011).
1.0 1
5.4. Summary of Extended Sensitivity Analysis for ENGP Tanker and Terminal
Operations
Our extended sensitivity analysis for ENGP tanker, terminal, and pipeline operations
assesses the impact of some deficiencies in the methodological approach used in the
ENGP application. We address DNV’s narrow scope of parameters for its sensitivity
analysis, illustrate the impact of a broader range of parameters on spill return periods,
and evaluate the sensitivity of mitigated tanker spills, unmitigated and mitigated
terminal spills, and pipeline spills. Therefore our extended sensitivity analysis
provides a more comprehensive evaluation of the effects of changes in parameters to
return periods for ENGP spills. Figure 3 presents a summary of our sensitivity analysis
and compares our results with the sensitivity analysis completed in the ENGP
application.
Overall, DNV’s sensitivity analysis for tanker spills represents small, incremental
changes to some of its input parameters that result in small changes to spill return
periods. Combined, sensitivity parameters tested by DNV reduce the unmitigated and
mitigated return period for oil/condensate spills to 56 years and 171 years,
respectively. Alternatively, our sensitivity analysis evaluates changes to key input
parameters that result in significant reductions to return periods for unmitigated and
mitigated spills of 23 and 34 years, respectively (Figure 3). Significant decreases in
return periods as a result of our extended sensitivity analysis illustrate the
considerable impact that changes in assumptions have on spill return periods and
show the importance of providing transparent documentation justifying assumptions
as well as the need for more extensive sensitivity analysis to communicate
uncertainties associated with forecasting spills.
Although Enbridge and its consultants fail to complete a sensitivity analysis for
terminal and pipeline spills, our sensitivity analysis demonstrates the impact of
different assumptions on spill return periods for these operations. For terminal spills,
our sensitivity analysis results in an unmitigated spill return period of 15 to 19 years,
which compares to the return period of 29 years estimated by DNV for any size
oil/condensate terminal spill. For pipeline spills, our sensitivity analysis uses different
data inputs based on historical data from Enbridge and the NEB and shows that a
pipeline spill could occur at least once per year. These pipeline spill return periods
53

epresent a considerable increase in spill likelihood compared to estimates from
WorleyParsons (2012) of one oil/condensate pipeline leak every two years.
Figure 3: Summary of Sensitivity Analysis for ENGP Tanker, Terminal, and Pipeline Spills
Note: All return periods represent oil/condensate spills
Source: Based on calculations from Brandsæter and Hoffman (2010); Enbridge (2010b Vol. 7B; 2012); NEB (2011).
6. Application of Oil Spill Risk Analysis Model to the ENGP
The United States developed a comprehensive risk assessment model that it has used to
evaluate oil spill risks for US oil and gas exploration and development for several decades.
Given the widespread use and acceptance of the US OSRA model, it is useful to apply it to
the ENGP to assess spill risk. The following section describes this model and applies it to
generate spill probability estimates for ENGP tanker and pipeline operations. The US
model is then evaluated with best practice guidelines for risk assessment in order to
identify any major deficiencies of the approach to estimating spill probabilities for the
ENGP.
6.1. Overview of Oil Spill Risk Analysis Model
The OSRA model was developed in 1975 by the US Department of the Interior (Smith et
al. 1982) and is a well-­‐established methodology that the US government uses to
estimate oil spill occurrence probabilities, as well as the chances of spills contacting
environmental resources. Although there is no legislative requirement to use the OSRA
model, concerns related to oil spill prevention and contingency planning following
implementation of the Oil Pollution Act of 1990 (U.S. Public Law 101-­‐380, August 18,
1990) have created a sustained interest in estimating oil spill probabilities among all
stakeholder groups (Anderson and LaBelle 2000). The Bureau of Ocean Energy
Management (BOEM), formerly the Minerals Management Service, examines
commercial oil and gas development on the Outer Continental Shelf adjacent to the
54

United States of America with the OSRA model (Price et al. 2003). The OSRA model is
also used by the BOEM to prepare environmental documents pursuant to the National
Environmental Policy Act, as well as for environmental assessments, consultations for
endangered species and fish habitat, oil spill planning and preparedness by the Bureau
of Safety and Environmental Enforcement (BSEE) (Anderson et al. 2012), and other
federal and state government agencies including the US Fish and Wildlife Service, the
National Marine Fisheries Service, and the US Coast Guard (Price et al. 2003).
Furthermore, industry uses the OSRA model in the preparation of oil spill contingency
plans (Anderson et al. 2012).
The OSRA model has been used in several specific applications in the US Outer
Continental Shelf. BOEM typically prepares an environmental impact statement for
offshore lease areas and a component of the environmental impact statement is an
evaluation of potential oil spill risk over the life of the projects under consideration
(Anderson and LaBelle 1990). The US federal government used the OSRA model to
examine spill probabilities associated with the development of offshore resources in
the Beaufort Sea in northern Alaska (US DOI 1997), Cook Inlet in southern Alaska (US
DOI 2003b), and the Gulf of Mexico (US DOI 2002; 2007b), and examined development
on the Pacific Outer Continental Shelf off the coast of California (US DOI 2000a), oil and
gas development on the Eastern Gulf of Mexico Outer Continental Shelf (US DOI 1998),
and other development and exploration activities in the Beaufort Sea (US DOI 2000b;
US DOI 2007a). The model continues to be used in US offshore leasing areas at present.
The OSRA model consists of three main components: (1) probability of spills occurring;
(2) trajectory simulation of spills, and; (3) combined spill probabilities and trajectory
simulations (Smith et al. 1982). The probability assessment developed by the US
Department of the Interior uses a per-­‐volume methodology that relies on historical
spill occurrences and volume of oil handled. A fundamental component of this
forecasting method is defining an appropriate exposure variable, which is defined as a
“…quantity related to oil production or transportation which has a precise statistical
relationship to spill occurrence” (Smith et al. 1982). Smith et al. (1982) assert that spill
occurrence estimates “depend fundamentally on the estimated amount of oil to be
produced” (p. 20) and recommend volume due to the variable’s simplicity and
predictability. Similarly Lanfear and Amstutz (1983) suggest that the volume of oil
handled is “… the most practical exposure variable for predicting oil spill occurrences
as a Poisson process” (p. 359). More recently, Anderson et al. (2012) and Anderson
and LaBelle (2000) support Smith et al. (1982) and Lanfear and Amstutz (1983) in the
recommendation of volume as the exposure variable in the OSRA because it satisfies
two important criteria: (a) volume is easy to define, and; (b) volume is a quantity that
can be estimated based on historical volumes of oil handled 26 .
26 For pipeline spills, Eschenbach et al. (2010) suggest that pipeline-­‐mile years is more directly linked to the
probability of oil spills compared to production volume and reason that the probability of a spill from a
network of pipelines depends on the length of the pipeline. However, Eschenbach et al. also determine that
production volume is an appropriate exposure variable since it passes a goodness of fit test.
55

The OSRA assumes that spills occur independently of each other as a Poisson process 27 .
Spill occurrence conforms to a Poisson process for three reasons: (1) no spill can occur
when the volume of oil produced or transported is equal to zero; (2) the record shows
that spill events are independent of each other over time and volume, and; (3) the
number of events in any interval are Poisson distributed and this process has
stationary increments (Anderson et al. 2012; Anderson and LaBelle 2000). Smith et al.
(1982) describe the probability (P) of a specific number of spills (n) in the course of
handling t barrels where λ is the rate of spill occurrence:
Equation 1: OSRA Model Equation (Poisson Assumption)
The above formula bases the rate at which spills occur (λ), which is typically expressed
as the mean number of spills per billion barrels (Bbbl) of oil handled, on historic spill
occurrence data and the volume of oil produced/transported (Anderson et al. 2012;
Anderson and LaBelle 2000). The BSEE collects tanker spill data from the United
States Coast Guard and from international sources, and manages a database that
provides spill rate data for tanker spills that occur at sea and in port (Anderson et al.
2012). Tanker spill rates are also based on data from the US Department of Commerce,
the US Army Corps of Engineers, and British Petroleum’s Statistical Review of World
Energy (Anderson et al. 2012). Anderson et al. (2012) and Anderson and LaBelle
(2000) estimate tanker spill rates for spills ≥ 1,000 bbl (159 m 3 ) because spills of this
size are more likely to be reported compared to smaller spills and the historical data
are considered more comprehensive than data for spills less than 1,000 bbl. Anderson
and LaBelle (2000) estimate spill rates for pipeline systems in Alaska with historical
data from the Minerals Management Service and the State of Alaska. Spill rates
developed by Anderson et al. (2012) and Anderson and LaBelle (1990; 1994; 2000)
have been regularly updated and improved to reflect changes in spill rates over time.
6.2. Spill Probability Estimates for ENGP Tanker Operations
We develop probability estimates for oil and condensate spills for ENGP tanker
operations based on the OSRA model 28 . The approach to estimating spill probabilities
with the OSRA model includes four main steps:
1. Obtain historical spill rate data
27 Smith et al. (1982) recognize that this assumption can be questioned particularly in the case that safety
standards are improved after a particular spill (i.e. introduction of double hull tankers). However, Smith et al.
acknowledge that there is sufficient evidence that a Poisson process provides a reasonable approximation for
spill occurrence (Stewart and Kennedy 1978 as cited in Smith et al. 1982).
28 According to Anderson et al. (2012), spill rates for tanker incidents in international waters and Valdez,
Alaska represent crude oil tanker spills. We use crude oil tanker spill rates to approximate condensate tanker
spill rates because the condensate-­‐carrying vessels are similar to vessels carrying crude oil and specific
incident data for condensate tankers are insufficient. We note that this approach is consistent with that used
by DNV in the ENGP regulatory application.
56

2. Estimate the volume of oil and condensate handled over the operational life of
the project being assessed (ENGP)
3. Calculate spill probabilities with the OSRA model in Equation 1 based on data
from steps (1) and (2)
4. Complete a sensitivity analysis to examine the effect of changes in inputs to spill
probabilities.
Anderson et al. (2012) provide the most recent historical spill rate data for use in the
OSRA model for international tanker spills and spills from tankers transporting Alaska
North Slope crude oil. As shown in Table 37, spill rates determined by Anderson et al.
(2012) range from 0.32 to 0.84 spills per Bbbl for tanker spills ≥ 1,000 bbl (159 m 3 )
and range from 0.11 to 0.42 spills per Bbbl for spills ≥ 10,000 bbl (1,590 m 3 ).
International spill rates for the largest spill size category identified by Anderson et al.
(2012) range between 0.03 and 0.17 for spills ≥ 100,000 bbl (15,900 m 3 ). Anderson et
al. (2012) suggest that regulatory changes in the 1990s requiring double hull tankers
are the likely cause of continued declines in tanker spills over time.
Table 37: International and Alaska Oil Tanker Spill Rates
Spill size Source Data
Time
Period
Number of
Spills
Volume
Transported
(Bbbl)
Spill Rate
(per Bbbl)
≥ 1,000 bbl
(≥ 159 m3) International
Valdez, Alaska
1974-­‐2008
1994-­‐2008
1977-­‐2008
1989-­‐2008
303
59
11
4
359.9
185.8
15.3
8.7
0.84
0.32
0.72
0.46
Range of Spill Rates (per Bbbl) 0.32 -­‐ 0.84
≥ 10,000 bbl
(≥ 1,590 m3) International
Valdez, Alaska
1974-­‐2008
1994-­‐2008
1977-­‐2008
1989-­‐2008
151
20
3
1
359.9
185.8
15.3
8.7
0.42
0.11
0.20
0.12
Range of Spill Rates (per Bbbl) 0.11 -­‐ 0.42
≥ 100,000 bbl
(≥ 15,900 m3) International
1974-­‐2008
1994-­‐2008
62 359.9
6 185.8
Range of Spill Rates (per Bbbl)
0.17
0.03
0.03 -­‐ 0.17
Source: Anderson et al. (2012).
Note: Bbbl represents one billion barrels of oil.
In the second step, we estimate the volume of oil and condensate transported by ENGP
tanker traffic over the operational life of the project. The amount of crude oil
potentially shipped out of Kitimat Terminal is equal to pipeline capacity of 525 kbpd,
which equates to approximately 191.6 million barrels per year. Over a 30-­‐year and 50-­‐
year operating period, a volume of 525 kbpd represents over 5.7 Bbbl and 9.6 Bbbl,
respectively, of oil transported by tankers. The volume of condensate is equal to
pipeline capacity of 193 kbpd or 70.4 million barrels per year. Condensate throughput
over a 30-­‐year and 50 year period equals 2.1 and 3.5 Bbbl, respectively. These oil and
condensate volumes represent the baseline throughput capacities in the ENGP
regulatory application.
57

In the third step, we use spill rates for tanker spills and the volume of oil and
condensate transported by ENGP tanker traffic as inputs to the OSRA model (Equation 1)
and estimate spill probabilities. Our application of the OSRA model uses the following
four spill occurrence rates to determine tanker spill probabilities for the ENGP:
i. Spills in international waters between 1974 and 2008, which represent the
entire record of worldwide tanker spills from the BOEM database
ii. Spills in international waters between 1994 and 2008, which represent
worldwide spills from modern tanker operations in the last 15 years of the
BOEM database
iii. Spills associated with shipments departing Valdez, Alaska between 1977 and
2008, which represent the entire record of tanker spills transporting Alaska
North Slope crude oil
iv. Spills associated with shipments departing Valdez, Alaska between 1989 and
2008 that represent the last 20 years of modern tanker traffic transporting
Alaskan crude oil.
As shown in Table 38, probabilities for ENGP tanker spills are high over a 30-­‐year period
at baseline throughput capacities. According to the OSRA model, spill probability for
ENGP tanker operations ranges from 84.1% to 99.2% for oil spills ≥ 1,000 bbl, and spill
probability increases to between 91.9% and 99.9% for an oil or condensate spill ≥
1,000 bbl. Probabilities for larger ENGP oil/condensate tanker spills range between
57.9% and 96.3% for spills ≥ 10,000 bbl, and between 21.0% and 73.7% for spills ≥
100,000 bbl over a 30-­‐year period. The lower probabilities are based on the 1994 to
2008 international data and given improvements in safety, these lower probabilities
are likely more indicative of future spill rates.
Table 38: ENGP Tanker Spill Probabilities based on OSRA Model (30 years)
Spill size Source Data
Time
Period
Oil
30-­‐Year Spill Probability (%)
Condensate Combined
≥ 1,000 bbl
(≥ 159 m3) International
Valdez, Alaska
1974-­‐2008
1994-­‐2008
1977-­‐2008
1989-­‐2008
99.2
84.1
98.4
92.9
83.1
49.1
78.2
62.2
99.9
91.9
99.7
97.3
Range 84.1 -­‐ 99.2 49.1 -­‐ 83.1 91.9 -­‐ 99.9
≥ 10,000 bbl
(≥ 1,590 m3) International
Valdez, Alaska
1974-­‐2008
1994-­‐2008
1977-­‐2008
1989-­‐2008
91.1
46.9
68.3
49.8
58.8
20.7
34.5
22.4
96.3
57.9
79.2
61.1
Range 46.9 -­‐ 91.1 20.7 -­‐ 58.8 57.9 -­‐ 96.3
≥ 100,000 bbl
(≥ 15,900 m3) International
1974-­‐2008
1994-­‐2008
Range
62.4
15.8
15.8 -­‐ 62.4
30.2
6.1
6.1 -­‐ 30.2
73.7
21.0
21.0 -­‐ 73.7
Source: Calculated based on Anderson et al. (2012).
Over a 50 year operating period, the minimum probability for oil spills ≥ 1,000 bbl is
95.3%, which increases to 98.5% for an oil or condensate tanker spill (Table 39). The
respective minimum probability for oil/condensate spills ≥ 10,000 bbl and ≥ 100,000
bbl is 76.3% and 32.5%, respectively. Again, the lower probabilities are based on the
58

1994 to 2008 international data and given improvements in safety, these lower
probabilities are likely more indicative of future spill rates.
Table 39: ENGP Tanker Spill Probabilities based on OSRA Model (50 years)
Spill size Source Data
Time
Period
Oil
50-­‐Year Spill Probability (%)
Condensate Combined
≥ 1,000 bbl
(≥ 159 m3) International
Valdez, Alaska
1974-­‐2008
1994-­‐2008
1977-­‐2008
1989-­‐2008
99.9
95.3
99.9
98.8
94.8
67.6
92.1
80.2
99.9
98.5
99.9
99.8
Range 95.3 -­‐ 99.9 67.6 -­‐ 94.8 98.5 -­‐ 99.9
≥ 10,000 bbl
(≥ 1,590 m3) International
Valdez, Alaska
1974-­‐2008
1994-­‐2008
1977-­‐2008
1989-­‐2008
98.2
65.1
85.3
68.3
77.2
32.1
50.6
34.5
99.6
76.3
92.7
79.2
Range 65.1 -­‐ 98.2 32.1 -­‐ 77.2 76.3 -­‐ 99.6
≥ 100,000 bbl
(≥ 15,900 m3) International
1974-­‐2008
1994-­‐2008
Range
80.4
25.0
25.0 -­‐ 80.4
45.1
10.0
10.0 -­‐ 45.1
89.2
32.5
32.5 -­‐ 89.2
Source: Calculated based on Anderson et al. (2012).
In the fourth step, we complete a sensitivity analysis for the OSRA model by increasing
the volume of oil transported from 525 kbpd to 850 kbpd and increasing the volume of
condensate from 193 kbpd to 275 kbpd as outlined by Enbridge in its response to an
information request (Enbridge 2010b Information Request No. 3). The OSRA model
estimates a minimum probability of 94.9% that an oil tanker spill ≥ 1,000 bbl could
occur over a 30-­‐year period at a throughput of 850 kbpd of oil. The minimum
probability increases to 98.1% for an oil or condensate spill ≥ 1,000 bbl at higher
throughput capacities of 850 kbpd of oil and 275 kbpd of condensate (Table 40). The
minimum probability for an oil/condensate spill ≥ 10,000 bbl is 74.2% in a 30-­‐year
period suggesting a spill of this size could likely occur over the life of the ENGP.
Table 40: Sensitivity Analysis for ENGP Tanker Spill Probabilities at Higher Capacities (30 Years)
Spill size Source Data
Time
Period
Oil
30-­‐Year Spill Probability (%)
Condensate Combined
≥ 1,000 bbl
(≥ 159 m3) International
Valdez, Alaska
1974-­‐2008
1994-­‐2008
1977-­‐2008
1989-­‐2008
99.9
94.9
99.9
98.6
92.0
61.8
88.6
75.0
99.9
98.1
99.9
99.7
Range 94.9 -­‐ 99.9 61.8 -­‐ 92.0 98.1 -­‐ 99.9
≥ 10,000 bbl
(≥ 1,590 m3) International
Valdez, Alaska
1974-­‐2008
1994-­‐2008
1977-­‐2008
1989-­‐2008
98.0
64.1
84.5
67.3
71.8
28.2
45.2
30.3
99.4
74.2
91.5
77.2
Range 64.1 -­‐ 98.0 28.2 -­‐ 71.8 74.2 -­‐ 99.4
≥ 100,000 bbl
(≥ 15,900 m3) International
1974-­‐2008
1994-­‐2008
Range
79.4
24.4
24.4 -­‐ 79.4
40.1
8.6
8.6 -­‐ 40.1
87.7
30.9
30.9 -­‐ 87.7
Source: Calculated based on Anderson et al. (2012).
59

Spill probabilities for an ENGP oil/condensate tanker spill increase significantly over a
50-­‐year operating period at higher throughput capacities (Table 41). Indeed, the OSRA
model estimates a high likelihood (89.5%) that an oil/condensate spill ≥ 10,000 bbl
could occur over the life of the project. For larger spills, The OSRA model estimates
there is at least a 46.0% chance that an oil/condensate spill nearly half the size of the
Exxon Valdez oil spill of 41,000 m 3 could occur within a 50-­‐year period of the ENGP
operating at higher throughput capacities.
Table 41: Sensitivity Analysis for ENGP Tanker Spill Probabilities at Higher Capacities (50 years)
Spill size Source Data
Time
Period
Oil
50-­‐Year Spill Probability (%)
Condensate Combined
≥ 1,000 bbl
(≥ 159 m3) International
Valdez, Alaska
1974-­‐2008
1994-­‐2008
1977-­‐2008
1989-­‐2008
99.9
99.3
99.9
99.9
98.5
79.9
97.3
90.1
99.9
99.9
99.9
99.9
Range 99.3 -­‐ 99.9 79.9 -­‐ 98.5 99.9
≥ 10,000 bbl
(≥ 1,590 m3) International
Valdez, Alaska
1974-­‐2008
1994-­‐2008
1977-­‐2008
1989-­‐2008
99.9
81.8
95.5
84.5
87.9
42.4
63.3
45.2
99.9
89.5
98.4
91.5
Range 81.8 -­‐ 99.9 42.4 -­‐ 87.9 89.5 -­‐ 99.9
≥ 100,000 bbl
(≥ 15,900 m3) International
1974-­‐2008
1994-­‐2008
Range
92.8
37.2
37.2 -­‐ 92.8
57.4
14.0
14.0 -­‐ 57.4
97.0
46.0
46.0 -­‐ 97.0
Source: Calculated based on Anderson et al. (2012).
We note several limitations in our application of the OSRA model to the ENGP. The
principal limitation relates to the model’s reliance on historical data to estimate future
spill rates. As noted, there have been improvements in safety that resulted in a
reduction in spill rates over time (Anderson et al. 2012; Anderson and LaBelle 2000)
and thus historical rates may not accurately reflect future risk. Tanker spill rates in the
future could continue to decline as a result of further mitigation measures and
regulatory requirements, remain static due to diminishing returns from previous
regulatory and technological improvements, or increase as a result of potential future
deregulation or an aging tanker fleet. We do not attempt to predict future tanker spill
rates over the operational life of the ENGP. Second, average historical data may not
reflect the unique characteristics of the project environment that may affect risk such
as unusual hazards or special mitigation measures. We attempt to mitigate this
weakness by using tanker spill data from Alaska, which may increase the likelihood
that data more accurately reflect the unique risks of the Pacific west coast region.
Furthermore, Alaska data for the 1989 to 2008 period represents mitigation measures
similar to those proposed for the ENGP such as escort tugs. Third, Anderson et al.
(2012) do not estimate confidence intervals for tanker spill rates. Although Anderson
and LaBelle (2000) determine 95% confidence intervals for spill rates, the recent
update by Anderson et al. (2012) only provides the mean estimate for spill rates. We
attempt to address this issue by providing a range of spill probabilities based on
different spill occurrence rates.
60

6.3. Spill Probability Estimates for ENGP Pipeline
We also develop probability estimates for ENGP pipeline spills with the OSRA model.
Our approach to estimating pipeline spill probabilities with the OSRA model follows
the same approach as that used to determine probabilities for tanker spills.
In the first step, we obtain spill rates from Anderson and LaBelle (2000) based on
historical data for the Trans-­‐Alaska Pipeline System and onshore Alaska North Slope
crude oil production (Table 42). Between 1977 and 1998, six spills ≥ 500 bbl (80 m 3 )
and five spills ≥ 1,000 bbl (159 m 3 ) occurred from the Trans-­‐Alaska Pipeline System,
resulting in a spill rate of 0.48 and 0.40, respectively, for every Bbbl moved. A single
spill ≥ 500 bbl (80 m 3 ) associated with the Trans-­‐Alaska Pipeline System occurred
between 1985 and 1998 producing a spill rate of 0.12. Anderson and LaBelle (2000)
also determine that a single Alaskan North Slope crude oil and condensate spill ≥ 500
bbl occurred from onshore pipelines between 1985 and 1998, yielding a spill rate
similar to the Trans-­‐Alaska Pipeline System rate of 0.12 spills for every Bbbl
transported.
Table 42: Spill Rates for Pipeline Spills in Alaska based on Anderson and LaBelle (2000)
Spill Size Source Data
≥ 500 bbl
(80 m 3)
≥ 1,000 bbl
(159 m 3)
Trans-­‐Alaska Pipeline System
Time Number of Spill Rate
Period Spills (per Bbbl)
1977-­‐1998 6 0.48
1985-­‐1998 1 0.12
Alaska North Slope 1985-­‐1998 1 0.12
Trans-­‐Alaska Pipeline System
1977-­‐1998 5 0.40
1985-­‐1998 0 -­‐
Alaska North Slope 1985-­‐1998 0 -­‐
Source: Anderson and LaBelle (2000).
In the second step, we estimate the volume of oil and condensate transported by the
Northern Gateway Pipeline over an operating period of 30 to 50 years. Since pipeline
spill rates from Anderson and LaBelle (2000) represent oil and condensate, we assume
pipeline throughput of 718 kbpd for the ENGP based on oil pipeline volume of 525
kbpd and condensate volume of 193 kbpd. Combined oil and condensate throughput
represents 7.9 Bbbl over 30 years and 13.1 Bbbl over 50 years. We also complete a
sensitivity analysis examining the effects of increased capacity to 1,125 kbpd (850
kbpd of oil and 275 kbpd of condensate), which equates to approximately 12.3 Bbbl
over 30 years and 20.5 Bbbl over 50 years.
In the third step, we use the following two spill occurrence rates from Anderson and
LaBelle (2000) to determine a range of probabilities for ENGP pipeline spills ≥ 500 bbl:
i. Spills from the Trans-­‐Alaska Pipeline System between 1977 and 1998, which
represent the entire record of pipeline spills from the BOEM database
ii. Spills from the Trans-­‐Alaska Pipeline System and Alaska North Slope pipelines
between 1985 and 1998, which represent the most recent record of pipeline
spills provided by BOEM.
61

We input the spill rates for pipeline spills ≥ 500 bbl and the volume of oil transported
by the ENGP pipeline in Equation 1 to estimate probabilities for ENGP pipeline spills
(Table 43). Over a 30-­‐year operating period at planned throughput capacity, the
probability of an ENGP pipeline spill ≥ 500 bbl ranges between 61.1% and 97.7% and
the probability increases to between 79.2% and 99.8% over a 50-­‐year operating
period.
Table 43: Pipeline Spill Probabilities based on OSRA Model (30 and 50 Years)
Spill Size Source Data
Time
Period
30-­‐Year Spill
Probability (%)
50-­‐Year Spill
Probability (%)
≥ 500 bbl
(80 m3) Trans-­‐Alaska Pipeline System 1977-­‐1998 97.7 99.8
Trans-­‐Alaska Pipeline System /
Alaska North Slope
1985-­‐1998 61.1 79.2
Range 61.1 -­‐ 97.7 79.2 -­‐ 99.8
Source: Calculated based on Anderson and LaBelle (2000).
Our sensitivity analysis examines changes in spill probabilities if the volume of oil
transported by the pipeline increases 62% to 850 kbpd and the volume of condensate
increases 42% to 275 kbpd. Over 30-­‐ and 50-­‐year operating periods, the minimum
likelihood of a spill ≥ 500 bbl increases to 77.2% and 91.5%, respectively, when
pipeline throughput increases to 1,125 kbpd.
Table 44: Sensitivity Analysis for Pipeline Spill Probabilities at Higher Capacities (30 and 50
Years)
Spill Size Source Data
Time
Period
30-­‐Year Spill
Probability (%)
50-­‐Year Spill
Probability (%)
≥ 500 bbl
(80 m3) Trans-­‐Alaska Pipeline System 1977-­‐1998 99.7 99.9
Trans-­‐Alaska Pipeline System /
Alaska North Slope
1985-­‐1998 77.2 91.5
Range 77.2 -­‐ 99.7 91.5 -­‐ 99.9
Source: Calculated based on Anderson and LaBelle (2000).
The application of the OSRA model with spill rates from pipeline spills in Alaska to the
ENGP has several limitations. First, it assumes that historical pipeline spill data from
Alaska is appropriate to estimate future spills for ENGP pipelines, which may have
different risks due to different technological, safety, and regulatory standards. The
counter argument to this limitation is that available historical spill data from
Enbridge’s own liquids pipeline system does not indicate improvements in pipeline
safety between 1998 and 2010 and thus there is no rationale for reducing spill rates.
Further, a study of spill rates in the US from Eschenbach et al. (2010) determined no
significant decline in pipeline spill rates between 1972 and 2005. Another limitation is
that the OSRA pipeline data only includes spills ≥ 500 bbl, and therefore excludes a
large proportion of smaller spills (i.e. spills less than 500 bbl) that may cause
environmental, economic, and sociocultural damage. Finally, Anderson and LaBelle
(2000) do not provide confidence intervals for pipeline spills.
62

6.4. Summary of ENGP Spill Probabilities Estimated with OSRA Model
The summary of spill probabilities for ENGP tanker and pipeline spills calculated with
the OSRA model in Table 45 shows that oil/condensate spill probabilities for the ENGP
are high, particularly for spills ≥ 1,000 bbl. Indeed, the OSRA model estimates that the
minimum probability of a tanker spill ≥ 1,000 bbl over the 30-­‐ and 50-­‐year life of the
ENGP is 91.9% and 98.5%, respectively. For spills ≥ 10,000 bbl, the minimum
probability of a spill over the life of the ENGP is 57.9%, which increases to 89.5% when
the ENGP operates over a 50-­‐year period at higher throughput capacities. Based on the
OSRA, there is at least a 32.5% chance that a spill nearly half the size of the Exxon
Valdez oil spill of 41,000 m 3 could occur within a 50-­‐year period of the ENGP operating
at its planned throughput capacity and the minimum probability of a spill increases to
46.0% if the ENGP ships higher volumes of oil and condensate over a 50-­‐year period.
Table 45 also presents an overall spill probability for the entire ENGP, which ranges
between 96.8% and 99.9% over the life of the project 29 . Thus, the OSRA model
estimates that there is a very high likelihood that either a tanker spill ≥ 1,000 bbl or a
pipeline spill ≥ 500 bbl could occur during ENGP operations.
Table 45: Summary of Tanker and Pipeline Oil/Condensate Spill Probabilities for the ENGP
Type and Size Spill
Tanker Spill
30-­‐Year Spill Probability (%)
718 kbpd* 1,125 kbpd**
50-­‐Year Spill Probability (%)
718 kbpd* 1,125 kbpd**
≥ 1,000 bbl (≥ 159 m3) 91.9 -­‐ 99.9 98.1 -­‐ 99.9 98.5 -­‐ 99.9 99.9
≥ 10,000 bbl (≥ 1,590 m3) 57.9 -­‐ 96.3 74.2 -­‐ 99.4 76.3 -­‐ 99.6 89.5 -­‐ 99.9
≥ 100,000 bbl (≥ 15,900 m3) Pipeline Spill*
21.0 -­‐ 73.7 30.9 -­‐ 87.7 32.5 -­‐ 89.2 46.0 -­‐ 97.0
≥ 500 bbl (≥ 80 m3) Tanker or Pipeline Spill
61.1 -­‐ 97.7 77.2 -­‐ 99.7 79.2 -­‐ 99.8 91.5 -­‐ 99.9
See Note 96.8 -­‐ 99.9 99.6 -­‐ 99.9 99.7 -­‐ 99.9 99.9
* Represents oil volume of 525 kbpd and condensate volume of 193 kbpd (total of 718 kbpd),
** Represents oil volume of 850 kbpd and condensate volume of 275 kbpd (total of 1,125 kbpd).
Note: Combined tanker and pipeline spill calculated based on the following spill probabilities: (1) oil/condensate spill probabilities
for tanker spills ≥ 1,000 bbl and; (2) oil/condensate spill probabilities for pipeline spills ≥ 500 bbl.
6.5. Evaluation of OSRA Model with Best Practice Criteria for Risk Assessment
Similar to our evaluation of methods used to estimate spill return periods in the ENGP
regulatory application, we evaluate the OSRA method with best practice guidelines for
risk assessment (see Table 17). We qualitatively assess each criterion according to the
four-­‐point scale:
• Fully met: no deficiencies
• Largely met: no major deficiencies
• Partially met: at least one major deficiency
• Not met: two or more major deficiencies
29 We estimate overall spill probabilities for the ENGP based on: (1) oil/condensate spill probabilities for
tanker spills ≥ 1,000 bbl, and (2) oil/condensate spill probabilities for pipeline spills ≥ 500 bbl.
63

Transparency
Criterion: Documentation fully and effectively discloses supporting evidence, assumptions,
data gaps and limitations, as well as uncertainty in data and assumptions, and their
resulting potential implications to risk.
Our application of the OSRA model to the ENGP provides transparency. We clearly
summarize and provide supporting evidence for tanker and pipeline spill rates
determined by Anderson et al. (2012) and Anderson and LaBelle (2000), volumes for
oil and condensate over the operating life of the ENGP, and increases in the volume of
oil handled in the sensitivity analysis with reference to Enbridge’s phased expansion
that could increase the amount of oil handled from 525 kbpd to 850 kbpd and the
amount of condensate handled from 193 kbpd to 275 kbpd (Enbridge 2010b
Information Request No. 3). Furthermore, we transparently identify weaknesses of the
OSRA model applied to the ENGP, which consist of the model’s usage of historical data,
data from a different, albeit similar geographical area, and the model’s limited
availability of confidence intervals.
Evaluation: There are no major deficiencies related to the transparency criterion and
thus this criterion is fully met.
Replicability
Criterion: Documentation provides sufficient information to allow individuals other than
those who did the original analysis to obtain similar results.
Due to its parsimonious methodology, replicating the OSRA methodology is
straightforward as the model requires two major inputs: (1) the volume of oil handled,
and; (2) the spill rate based on historical spill occurrence data for various classes of
spills. In our application of the OSRA model, we estimate volume data for the ENGP
according to the planned volume of hydrocarbon transported in the ENGP regulatory
application. We derive spill rates from Anderson et al. (2012) and Anderson and
LaBelle (2000), and both studies disclose proprietary data required to replicate
historical spill rates for international tanker spills and tanker spills in US waters.
Furthermore, Smith et al. (1982) provide a detailed description of how to calculate spill
probabilities as a Poisson process for fixed classes of spills with volume of oil handled
as the exposure variable (see Equation 1).
Evaluation: There are no major deficiencies related to the replicability criterion and thus
this criterion is fully met.
Clarity
Criterion: Risk estimates are easy to understand and effectively communicate the nature
and magnitude of the risk in a manner that is complete, informative, and useful in
decision-­‐making.
64

The OSRA method clearly estimates spills as probability of occurrence over the life of
the project and probabilities represent various classes of tanker spills at port and at
sea, including smaller spills that, according to the US DOI (2003a) can have significant
adverse environmental effects (i.e. spills ≥ 1,000 bbl or ≥ 159 m 3 ). Furthermore, we
combine probability estimates for tanker spills and pipeline spills into a single value
that effectively communicates spill risk for the entire ENGP.
Evaluation: There are no major deficiencies related to the clarity criterion and thus this
criterion is fully met.
Reasonableness
Criterion: The analytical approach ensures quality, integrity, and objectivity, and meets
high scientific standards in terms of analytical methods, data, assumptions, logic, and
judgment.
The major deficiency in the OSRA method related to reasonableness is the model’s
reliance on average historical spill rates:
1. Historical spill data may not represent similar operating conditions as ENGP tanker
and pipeline operations
Several observations suggest that historical spill rate data from Anderson et al.
(2012) and Anderson and LaBelle (2000) may not reflect the unique
characteristics of the ENGP. First, average historical tanker spill data from
Anderson et al. (2012), which shows a reduction in spill frequency over time as a
likely result of regulatory changes requiring double hull tankers, may not reflect
future ENGP tanker operations. Future tanker operations for the ENGP could
include increased safety and technological measures such as escort tug operations,
enhanced navigation systems, different tanker characteristics and types, and
different regulatory standards that further reduce tanker spill rates over the
operational life of the project. However, the increasing age of the tanker fleet may
offset safety improvements, resulting in an increase in future spill rates. A recent
study from Eliopoulou et al. (2011) determines that non-­‐accidental structural
failure incident rates are higher for double-­‐hull tankers over the age of 15 years.
Thus historical spill rate data may underestimate future spill rates for particular
incident types such as non-­‐accidental structural failures, since these types of
accidents may become significant after 2020 (Papanikolaou et al. 2009). We avoid
any assumptions or speculation about future tanker spill rates. Second, historical
spill data may not reflect unique characteristics of the project environment that
could affect risk. International spill rate data from Anderson et al. (2012) based on
average tanker operations reflect a range of different operating conditions,
weather conditions, and maintenance programs, and thus these data may not
represent tanker traffic associated with the ENGP that would operate in a region
containing potential hazards related to maneuverability, visibility, and
meteorological and oceanographic conditions. Similarly, historical spill rate data
from the Trans-­‐Alaska Pipeline System and Alaska North Slope crude oil
65

transportation systems may not reflect the conditions of ecologically sensitive
regions that potentially present unique terrain and climactic conditions for the
ENGP oil and condensate pipeline. Third, the OSRA model does not consider any
increased risk associated with the corrosiveness of diluted bitumen. Given the
corrosive threat characteristic of higher sulphur concentrations (Lyons and Plisga
2005) and the higher sulphur content in dilbit (EC OPD 2012), as well as increased
internal corrosion rates observed between the Alberta and US pipeline systems,
the increased risk associated with transporting dilbit may increase spill occurrence
rates for tanker, terminal, and pipeline operations. We do not attempt to adjust
spill rate data from Anderson et al. (2012) or Anderson and LaBelle (2000) for any
of the aforementioned characteristics unique to ENGP tanker, terminal, and
pipeline operations. We note that despite these potential deficiencies, the OSRA
model has held up in US courts when it has been challenged (LaBelle 2012).
Evaluation: There is one major deficiency related to the reasonableness criterion and thus
this criterion is partially met.
Reliability
Criterion: Appropriate analytical methods explicitly describe and evaluate sources of
uncertainty and variability that affect risk, and estimate the magnitudes of uncertainties
and their effects on estimates of risk.
In the application of the OSRA model to the ENGP, we test the magnitude of
uncertainties with a sensitivity analysis of the two input parameters to the model. Our
sensitivity analysis examines the effects of an increase in volume of oil handled, and we
use various spill rates based on different historical data to provide a range of estimates
for spill probability. However, there is one major deficiency related to the reliability
criterion:
1. Incomplete inventory of confidence intervals for spill rates calculated based on
historical spill data
Anderson et al. (2012) and Anderson and LaBelle (2000) do not provide a
comprehensive inventory of confidence intervals for spill rates for use in the OSRA
model. The most recent spill rate data updated by Anderson et al. (2012) express
spill rates as the mean number of spills per Bbbl and omit confidence levels for
tanker spills at port and at sea. Although an earlier study from Anderson and
LaBelle (2000) calculates confidence intervals for tanker spill rates for spills
≥1,000 bbl, the authors do not determine confidence intervals for any tanker spills
≥10,000 bbl, including spills in international waters and spills from the
transportation of Alaska North Slope crude oil. Furthermore, Anderson and
LaBelle (2000) do not provide confidence intervals for pipeline spill rates
calculated based on historical data for the Trans-­‐Alaska Pipeline System and crude
oil/condensate spills associated with onshore Alaska North Slope crude oil
production.
66

Evaluation: There is one major deficiency related to the clarity criterion and thus this
criterion is partially met.
Validity
Criterion: Independent third-­‐party experts review and validate findings of the risk
analysis to ensure credibility, quality, and integrity of the analysis.
The OSRA model was developed in 1975 by the United States Department of the
Interior (Smith et al. 1982) and the methodology has undergone independent peer-­‐
review on numerous occasions since its development. In addition to the recent study
from Anderson et al. (2012) estimating oil spill occurrence, spill rate assumptions for
the OSRA model have been independently peer-­‐reviewed on four separate occasions
(BOEM undated): three previous studies by Anderson and LaBelle (1990; 1994; 2000)
examine occurrence rates used in the analysis of accidental oil spills on the US Outer
Continental Shelf and a study by Lanfear and Amstutz (1983) examines cumulative
frequency distributions of oil spills and determined that volume of oil handled is the
most practical exposure variable for estimating oil spill probabilities. We acknowledge
that, although BOEM (undated) refers to “independently peer-­‐reviewed papers”,
authors that undertook the reviews (Anderson, LaBelle, Lanfear, and Amstutz) are
identified with the BOEM (Minerals Management Service) of the US Department of the
Interior. However, the reviews were published in journals and were subject to a blind
peer review process prior to publication.
To ensure the validity of our application of the OSRA model to the ENGP, we submitted
the findings of this report to Robert P. LaBelle, Science Advisor with BOEM in the US
Department of the Interior. Mr. LaBelle is an expert in offshore energy exploration and
development and has been working with the OSRA model as early as 1985. Prior to
being Science Advisor, Mr. LaBelle was the acting Deputy Director for Bureau of Safety
and Environmental Enforcement and was the Associate Director for Offshore Energy at
the Bureau of Ocean Energy Management, Regulation and Enforcement. Mr. LaBelle
independently verified our application of the OSRA model to the ENGP and validated
our results.
Evaluation: There are no major deficiencies related to the validity criterion and thus this
criterion is fully met.
Risk Acceptability
Criterion: Stakeholders identify and agree on acceptable levels of risk in a participatory,
open decision-­‐making process and assess risk relative to alternative means of meeting
project objectives.
Determining levels of risk acceptability is an exercise that must take place in the
context of a larger planning process. Any discussion of risk acceptability depends on
the planning or regulatory review process and the stakeholders actively involved in
that process. We avoid any speculation on how the OSRA model would affect a
67

discussion around risk acceptability in the ENGP regulatory review process. We also
avoid any comparison of historical cases where the OSRA may have been an input to a
decision-­‐making process where risk acceptability is discussed, since the outcome
provides little insight on risk acceptability in the ENGP regulatory review process.
Thus, we cannot determine risk acceptability in the absence of a participatory planning
process in which tolerable levels of risk are discussed among stakeholders.
Evaluation: The criterion for risk acceptability is not applicable.
6.6. Summary of Findings for Qualitative Assessment of OSRA Model
Our evaluation of the OSRA model with best practices for risk assessment determined
that the methodology incorporates several of the best practices (Table 46). Four of the
seven best practice criteria for risk assessment have no major deficiencies and thus are
fully met. Our analysis reveals a total of two major deficiencies in the application of the
OSRA model to estimate spill probabilities for ENGP tanker and pipeline spills and thus
the two criteria of reasonableness and reliability were only partially met. The criterion
for risk acceptability is not applicable since defining risk acceptability must occur in
the context of a planning or project review process.
Table 46: Results for the Evaluation of the OSRA Model with Best Practices for Risk Assessment
Criterion Major Deficiencies Result
Transparency
Documentation fully and effectively discloses supporting
evidence, assumptions, data gaps and limitations, as
well as uncertainty in data and assumptions, and their
resulting potential implications to risk
Replicability
Documentation provides sufficient information to allow
individuals other than those who did the original
analysis to obtain similar results
Clarity
Risk estimates are easy to understand and effectively
communicate the nature and magnitude of the risk in a
manner that is complete, informative, and useful in
decision making
Reasonableness
The analytical approach ensures quality, integrity, and
objectivity, and meets high scientific standards in terms
of analytical methods, data, assumptions, logic, and
judgment
Reliability
Appropriate analytical methods explicitly describe and
evaluate sources of uncertainty and variability that
affect risk, and estimate the magnitudes of uncertainties
and their effects on estimates of risk
Validity
Independent third-­‐party experts review and validate
findings of the risk analysis to ensure credibility, quality,
and integrity of the analysis
Risk Acceptability
Stakeholders identify and agree on acceptable levels of
risk in a participatory, open decision-­‐making process
and assess risk relative to alternative means of meeting
project objectives
Note: n/a = not applicable.
No major deficiencies
No major deficiencies
No major deficiencies
1. Historical spill data may not represent
similar operating conditions as ENGP
tanker and pipeline operations
2. Incomplete inventory of confidence
intervals for spill rates calculated based
on historical spill data
No major deficiencies
Risk acceptability cannot be determined
in the absence of a participatory planning
process in which levels of risk are
discussed among stakeholders
Fully
Met
Fully
Met
Fully
Met
Partially
Met
Partially
Met
Fully
Met
N/A
68

7. Comparison of Spill Estimates for the ENGP
The following section compares spill forecast methodologies from the ENGP regulatory
application with the OSRA Model. We compare both methods by describing the
methodologies and spill estimates of each approach, as well as summarizing the results of
our qualitative assessment. We then discuss the implications of the results of our analysis
on the likelihood criterion required under the CEAA.
There are major differences between the methods estimating spill likelihood in the ENGP
regulatory application and the OSRA model (Table 47). First, ENGP risk analyses estimate
spill likelihood based on the number of spills per distance travelled by tankers, spills per
tanker operation at the marine terminal, and spills per km of pipeline, whereas the OSRA
model uses a different methodology based on the volume of hydrocarbons transported.
Thus for tanker spills, DNV estimates spill likelihood for only the BC portion of the tanker
trip while the OSRA estimates spill likelihood for the entire trip. Second, ENGP risk
analyses use different datasets than the OSRA model for tanker and pipeline spills and
these different datasets represent different spill size categories. Any comparison between
spill likelihood estimates from the ENGP regulatory application and estimates from the
OSRA model should consider the disparate spill size categories unique to each
methodology. Third, the different methodologies require different data inputs in order to
estimate spill probabilities. The methodology for estimating tanker spill likelihood in the
ENGP application uses six major data inputs, whereas the OSRA model is a much simpler
methodology that relies on two straightforward inputs. Fourth, both methodologies
produce different outputs that communicate very different conceptions of spill risk. The
ENGP regulatory application presents return periods for tanker, terminal, and pipeline
spills that represent the amount of time between spill events, whereas the OSRA model
presents the probability of a spill occurring over the operational life of the project, which in
the case of the ENGP is 30 to 50 years.
69

Table 47: Comparison of Methodologies Estimating Spill Likelihood in ENGP Risk Analyses and the
OSRA Model
Method
Source
Data
Spill Size
Categories
Major Data
Inputs
ENGP Risk Analyses OSRA Model
• Tanker: Spills per voyage
• Terminal: Spills per tanker operation
• Pipeline: Spills per kilometer of pipe
• Tanker and Terminal: LRFP data; DNV data
• Pipeline: NEB data; Data from recently
constructed pipelines, US PHMSA Data
• Tanker
o Any size spill
o Oil spills > 5,000 m3 (> 31,500 bbl)
o Oil spills > 20,000 m3 (> 125,800 bbl)
o Oil spills > 40,000 m3 (> 251,600 bbl)
• Terminal
o Spills < 1,000 m3 (< 6,300 bbl)*
• Pipeline
o Leaks and Ruptures^
• Tanker/Terminal:
o Incident frequencies
o Local scaling factors
o Conditional probabilities
o Route distance
o Tanker distribution
o Mitigation measures
• Pipeline:
o Spill frequencies
o Length of the pipeline
o Improvement factors
• Spills per volume of hydrocarbon
transported
• Tanker (in port and at sea): BSEE database
• Pipeline: State of Alaska data
• Tanker
o ≥ 1,000 bbl (≥ 159 m 3)
o ≥ 10,000 bbl (≥ 1,590 m 3)
o ≥ 100,000 bbl (≥ 15,900 m 3)
• Pipeline
o ≥ 500 bbl (≥ 80 m 3)
• Spill rates
• Volume of hydrocarbon handled
Risk
• Unmitigated and mitigated return periods • Probabilities over the life of the project
Outputs
* DNV refers to small (< 10 m3) and medium (10 m3 to 1,000 m3) terminal spills as ‘any size spill’ even though larger spills (> 1,000 m3) are not included.
^ Average size oil or condensate pipeline leak is 94 m3, average size oil pipeline rupture is 2,238 m3 and average size condensate rupture
is 823 m3 (WorleyParsons 2012).
Table 48 includes a summary of spill estimates for tanker, terminal, and pipeline spills
developed from the various methodologies for the ENGP at its planned throughput
capacity. Based on spill estimates from ENGP risk analyses, a tanker, terminal, or pipeline
spill could occur every 2 years 30 . Since return periods in the ENGP application estimate the
amount of time between spills and not the chance of a spill over the life of the project, we
restate return periods as probabilities over the life of the ENGP 31 . The results show that a
tanker, terminal, or pipeline spill has a 99.9% chance of occurring during 30 years of
30 Return period calculated based on the following ENGP spill probabilities: (1) any size tanker spill; (2) any
size terminal spill, and; (3) pipeline leaks.
31 To restate spill likelihood, we convert return periods to annual spill probabilities and estimate spill
probabilities over the life of the project. We make no other adjustments to any of the data inputs in DNV’s
methodology; our analysis simply restates return periods as probabilities over the life of the ENGP.
Furthermore, we do not adjust for any deficiencies identified in the qualitative assessment that would likely
result in further increases in spill probability.
70

operation. Our application of the OSRA model also estimates high spill probabilities for
ENGP tanker and pipeline operations. Based on the results of the OSRA modelling, the
minimum probability for a tanker or pipeline spill over a 30-­‐year operating period is 96.8%
and this increases to 99.7% over 50 years. When we restate spill probabilities from the
OSRA model as return periods, a spill from ENGP tanker or pipeline operations could occur
every 3 to 9 years 32 . Note that any comparison of the overall return periods and spill
probabilities for ENGP risk analyses and the OSRA model in Table 48 must consider the
different spill size categories for each method.
Table 48: Comparison of Return Periods and Spill Probabilities for the ENGP
Return Period Spill Probability (%)
(in years) 30 years 50 years
Tanker any size spill
Oil
Oil/Condensate
110 -­‐ 350
78 -­‐ 250
8.2 -­‐ 24.0
11.3 -­‐ 32.1
13.3 -­‐ 36.7
18.2 -­‐ 47.5
Tanker spill ≥ 31,500 bbl Oil 200 -­‐ 550 5.3 -­‐ 14.0 8.7 -­‐ 22.2
Terminal any size spill
Oil 34 -­‐ 90 28.5 -­‐ 59.2 42.8 -­‐ 77.5
Method Size and Type of Spill
ENGP
Risk
Analyses #
OSRA
Model
Oil/Condensate 29 -­‐ 62 38.6 -­‐ 65.1 55.6 -­‐ 82.7
Pipeline leak (594 bbl)
Oil
Oil/Condensate
4
2
99.9
99.9
99.9
99.9
ENGP Spill* Oil/Condensate 2 99.9 99.9
Tanker spill ≥ 1,000 bbl
Oil
Oil/Condensate
7 -­‐ 17
5 -­‐ 12
84.1 -­‐ 99.2
91.9 -­‐ 99.9
95.3 -­‐ 99.9
98.5 -­‐ 99.9
Tanker spill ≥ 10,000 bbl
Oil 13 -­‐ 48 46.9 -­‐ 91.1 65.1 -­‐ 98.2
Tanker spill ≥ 100,000 bbl
Oil/Condensate 10 -­‐ 35 57.9 -­‐ 96.3 76.3 -­‐ 99.6
Oil 31 -­‐ 174 15.8 -­‐ 62.4 25.0 -­‐ 80.4
Oil/Condensate 23 -­‐ 128 21.0 -­‐ 73.7 32.5 -­‐ 89.2
Pipeline Spill ≥ 500 bbl Oil/Condensate 8 -­‐ 32 61.1 -­‐ 97.7 79.2 -­‐ 99.8
ENGP Spill** Oil/Condensate 3 -­‐ 9 96.8 -­‐ 99.9 99.7 -­‐ 99.9
Source: Brandsæter and Hoffman (2010); Calculations based on Anderson et al. (2012); WorleyParsons (2012).
Note: Data for both methodologies represent planned throughput capacity of 525 kbpd of oil and 193 kbpd of condensate;
* Overall spill probability for ENGP based on spill probabilities for any size tanker spill, any size terminal spill, and pipeline leaks.
** Calculation based on spill probabilities for tanker spills ≥ 1,000 bbl and pipeline spills ≥ 500 bbl.
# Range represents unmitigated and mitigated spill data with the exception of pipeline spills.
To address challenges of comparing spill estimates from the fundamentally different
methodologies used in the ENGP regulatory application and OSRA model, we developed an
evaluative framework of best practice criteria for risk assessment to evaluate each
approach. Our analysis examines how each methodological approach satisfies best practice
criteria and identifies major deficiencies of both methodologies. As shown in Table 49, our
analysis identifies a total of 28 major deficiencies in the methodological approach used in
the ENGP application to estimate spills from ENGP operations. As a result, all seven criteria
are not met. The OSRA model has only two major deficiencies and fully meets four of the
seven best practice criteria.
32 To estimate return periods for the entire ENGP with the OSRA model, we take the inverse of the following
annual spill probabilities: (1) oil/condensate tanker spills ≥ 1,000 bbl and; (2) oil/condensate pipeline spills ≥
500 bbl.
71

Table 49: Summary and Comparison of Results for the Qualitative Assessment
Best Practice
ENGP Application OSRA Model
Criterion for Risk
Major Qualitative Major
Qualitative
Assessment
Deficiencies Assessment Deficiencies Assessment
Transparency 6 Not Met 0 Fully Met
Replicability 5 Not Met 0 Fully Met
Clarity 5 Not Met 0 Fully Met
Reasonableness 5 Not Met 1 Partially Met
Reliability 2 Not Met 1 Partially Met
Validity 2 Not Met 0 Fully Met
Risk Acceptability 3 Not Met n/a n/a
Total
N/A = not applicable.
28 -­‐ 2 -­‐
Results of our qualitative assessment provide insight on the relative strengths and
weaknesses of the methodologies used in the ENGP application and the OSRA model.
Methodologies estimating spill risk in the ENGP application contain major deficiencies that
include a lack of transparency for several key data inputs that reduce spill likelihood, a
failure to communicate risk as the probability of a spill over the life of the project, no
consideration of risks from the project that might occur outside the study area, a lack of
expert review of main findings, and a failure to define acceptable levels of risk in terms of
stakeholder values. Based on these deficiencies, methodologies estimating spill likelihood
in the ENGP application fail to address the CEAA decision criterion.
Our qualitative assessment of the OSRA model suggests that the OSRA model meets more of
the best practices criteria than the ENGP application, which is partly attributable to the
parsimonious approach of the OSRA model. Our application of the OSRA model to the
ENGP has only two major deficiencies and fully meets four of the best practices for risk
assessment methodologies. The OSRA model presents spills as probability of occurrence
over the 30-­‐ and 50-­‐year life of the project and thus satisfies the CEAA decision criterion for
the likelihood of significant adverse environmental effects. Indeed, the OSRA model
demonstrates that there is a high likelihood that the ENGP will cause significant adverse
environmental effects. The OSRA model also considers tanker spills outside Canada since
the methodology for estimating tanker spill probabilities uses a per-­‐volume approach
rather than the distance-­‐based method used by DNV that only examines tanker spills
within the territorial waters of Canada.
In summary, both methods have strengths and weaknesses and without further analysis we
are unable to confirm which method provides a more accurate estimate of the likelihood of
spills for the ENGP. However, based on our analysis the OSRA model satisfies CEAA
decision criteria whereas ENGP risk analyses fail to provide decision makers with the
necessary information required to accurately assess the likelihood of significant adverse
environmental effects from a spill. Notwithstanding the differences in methodologies, both
the OSRA model and ENGP risk analyses determine that a spill from the ENGP is likely to
occur.
72

8. Conclusion
This report evaluates spill estimates for ENGP tanker, terminal, and pipeline operations.
We evaluate spill estimates in the ENGP regulatory application, apply the OSRA model to
estimate spill probabilities for the ENGP, and compare the results of both approaches using
a qualitative assessment based on best practices for risk assessment. From our analysis we
conclude:
1. The ENGP Application Contains Major Deficiencies that Underestimate Spill Likelihood
Enbridge’s spill risk analysis contains 28 major deficiencies. As a result of these
deficiencies, Enbridge underestimates the risk of the ENGP. Some of the key deficiencies
include:
• Failure to present the probabilities of spills over the operating life of the ENGP
• Failure to evaluate spill risks outside the narrowly defined BC study area
• Reliance on LRFP data that underreport tanker incidents by between 38 and 96%
• Failure to include the expansion capacity shipment volumes in the analysis
• Failure to provide confidence ranges of the estimates
• Failure to provide adequate sensitivity analysis
• Failure to justify the impact of proposed mitigation measures on spill likelihood
• Potential double counting of mitigation measures
• Failure to provide an overall estimate of spill likelihood for the entire ENGP
• Failure to disclose information and data supporting key assumptions that were used
to reduce spill risk estimates
• Failure to use other well accepted risk models such as the US OSRA model.
2. Deficiencies in the ENGP Application Fail to Address the CEAA Decision Criterion
Due to deficiencies in the oil spill risk assessment, Enbridge does not provide an accurate
assessment of the likelihood of significant adverse environmental impacts as required
under CEAA. Enbridge expresses the likelihood of a spill as return periods for separate
components of the project (tanker, terminal, and pipeline) rather than the probability of a
spill over the life of the project. Without estimates of the probability of a spill over the
operating life of the project, decision makers are unable to determine whether the project
meets the CEAA criterion of likelihood of adverse environmental effects. Furthermore, due
to the deficiencies in the spill risk methodology, Enbridge fails to adequately disclose the
extent of risk associated with the ENGP and the uncertainty in estimating risk.
3. Restating Enbridge’s Own Findings as Probabilities Over Project Life Shows that Spill
Likelihood is High
When findings in the ENGP application are restated as the probability of a spill over the
operational life of the ENGP instead of return periods, the conclusion based on Enbridge’s
own analysis is that the likelihood of a spill is high (Table 50). These spill probabilities based
on Enbridge’s analysis underestimate spill likelihood because they are based on
methodological deficiencies.
73

Table 50: Probabilities for ENGP Tanker, Terminal, and Pipeline Spills (50 Years)
Spill Type (Oil or Condensate)
Spill Probability
over 50 years (%)
Tanker Spill 18.2
Terminal Spill 55.6
Pipeline Spill 99.9
Combined Spills 99.9
4. Sensitivity Analysis Shows that the Likelihood of a Spill is High
Our extended sensitivity analysis for the ENGP demonstrates that there is a higher
likelihood of spill occurrence than reported in the ENGP application. For tanker spills, the
extended sensitivity analysis evaluates changes to key input parameters such as increasing
shipments to the ENGP design capacity and combining parameter changes.
a. The extended sensitivity analysis shows that based on Enbridge’s own data, the rate
of tanker spills could range between 0.3 and 2.2 spills over a 50-­‐year period, much
higher than Enbridge’s estimate of less than one spill (Figure 4). This range still
understates spill risk because it does not correct for all the deficiencies in
Enbridge’s methodology.
Figure 4: Number of ENGP Tanker Spills in 50 Years (Oil and Condensate)
b. The extended sensitivity analysis for terminal spills determines that the rate of
spills in a 50-­‐year period could range between 1.2 and 3.4 spills for any size
terminal spill, significantly higher than Enbridge’s estimate of less than one spill in
50 years (Figure 5).
74

Figure 5: Number of ENGP Terminal Spills in 50 Years (Oil and Condensate)
c. The extended sensitivity analysis using spill frequency data from Enbridge’s liquid
pipeline system results in an estimated 776 oil and condensate pipeline spills in 50
years (or between 15 to 16 spills per year), which is 31 times more frequent than
Enbridge’s estimate of 25 spills in 50 years (Figure 6).
Figure 6: Number of ENGP Pipeline Spills in 50 Years (Oil and Condensate)
5. The OSRA Model Shows the Probability of a Major Spill is High
The OSRA model is the model the US government uses to estimate spill risk. The OSRA
model estimates 4 to 10 tanker spills over 1,000 bbl (159 m 3 ) could occur in a 50-­‐year
period (Figure 4), which is 20 to 50 times more frequent than Enbridge’s estimate. Due to
mitigation measures, the lower end of the OSRA spill range is more likely a better indicator
of future risk. In terms of spill probability, the OSRA model estimates that an
oil/condensate tanker spill ≥ 1,000 bbl has a 98.5% to 99.9% chance of occurring over a
50-­‐year operating period, well above Enbridge’s estimate of 18.2% (Figure 7). The lower
estimate of 98.5% is likely a better indication of future spill risk due to mitigation
measures. The probability of a tanker spill increases to 99.9% if the volume of oil and
condensate handled increases to the ENGP expansion capacity of 1,125 kbpd.
75

Figure 7: Probabilities for ENGP Tanker Spills over 50 years (Oil or Condensate)
6. Other Considerations
Finally, two important considerations must be reconciled in the ENGP regulatory
application: risk acceptability to stakeholders, and viable alternatives to the ENGP that
reduce risk. Risk acceptability in the ENGP application relies on a technical definition of
risk. Acceptable risk however is a value judgment that should be based on the definition of
risk as defined by impacted stakeholders. Stakeholders’ definition of acceptable risk
therefore needs to be assessed and used in the ENGP decision. Second, Enbridge has not
identified nor compared viable alternatives that meet the stated purpose of the ENGP.
There are viable alternatives to shipping crude oil from the Western Canada Sedimentary
Basin to market that involve no risk from oil tanker spills (Gunton and Broadbent 2012a)
and thus risk from tanker spills can be eliminated. These alternatives should be evaluated
relative to the ENGP to assess the most cost effective transportation option.
7. Overall Conclusion
We acknowledge that estimating risk is challenging due to the many uncertainties involved
and that different assumptions and methodologies will produce different assessments of
risk. The key in risk assessment is to test a range of assumptions and methodological
approaches to provide a clear and accurate assessment of the range of likely outcomes so
that decision makers can fully judge risk. The conclusion of this report is that Enbridge’s
oil spill risk assessment contains methodological deficiencies and does not therefore
provide an accurate assessment of the degree of risk associated with the ENGP. The risk
assessment in this report also concludes that the ENGP has a very high likelihood of a spill
that may have significant adverse environmental effects.
76

References
Alberta Energy and Utilities Board (AEUB). (2007). Pipeline Performance in Alberta, 1990-­‐
2005. Retrieved March 4, 2012, from
http://www.ercb.ca/docs/Documents/reports/r2007-­‐A.pdf.
AMEC Americas Limited (AMEC). (2012). Report on Quantitative Geohazard Assessment
Proposed Northern Gateway Pipelines. In WorleyParsons Canada Services Limited
(WorleyParsons) (Ed.), Semi-­‐Quantitative Risk Assessment (pp. 142).
Anderson, C., & LaBelle, R. P. (1990). Estimated Occurrence Rates for Analysis of Accidental
Oil Spills on the US Outer Continental Shelf. Oil and Chemical Pollution, 6(1), 21-­‐35.
Anderson, C., & LaBelle, R. P. (1994). Comparative Occurrence Rates for Offshore Oil Spills.
Spill Science & Technology Bulletin, 1(2), 131-­‐141.
Anderson, C., & LaBelle, R. P. (2000). Update of Comparative Occurrence Rates for Offshore
Oil Spills. Spill Science & Technology Bulletin, 6(5-­‐6), 303-­‐321.
Anderson, C., Mayes, M. & LaBelle, R. (2012). Update of Occurrence Rates for Offshore Oil
Spills. Retrieved July 21, 2012, from
http://www.boem.gov/uploadedFiles/BOEM/Environmental_Stewardship/Environ
mental_Assessment/Oil_Spill_Modeling/AndersonMayesLabelle2012.pdf.
Bercha Group Limited (Bercha Group). (2012a). Northern Gateway Kitimat Terminal
Quantitative Risk Analysis. Retrieved 2012, June 12, from
http://gatewaypanel.review-­‐examen.gc.ca/clf-­‐nsi/pblcrgstr/pblcrgstr-­‐eng.html.
Bercha Group Limited (Bercha Group). (2012b). Northern Gateway Pipeline Public Safety
Quantitative Risk Analysis. Retrieved 2012, June 12, from
http://gatewaypanel.review-­‐examen.gc.ca/clf-­‐nsi/pblcrgstr/pblcrgstr-­‐eng.html.
Bercha Group Limited (Bercha Group). (2012c). Northern Gateway Pipeline Pump Station
Quantitative Risk Analysis. Retrieved 2012, June 12, http://gatewaypanel.review-­‐
examen.gc.ca/clf-­‐nsi/pblcrgstr/pblcrgstr-­‐eng.html.
Brandsæter, A., & Hoffman, P. (2010). Marine Shipping Quantitative Risk Analysis: Enbridge
Northern Gateway Project. Prepared by Det Norske Veritas. Retrieved from
http://gatewaypanel.review-­‐examen.gc.ca/clf-­‐nsi/pblcrgstr/pblcrgstr-­‐eng.html.
77

British Columbia Ministry of Environment, Land, and Parks (BC MoELP). (2000).
Environmental Risk Assessment (ERA): An Approach for Assessing and Report
Environmental Conditions. (No. QH541.15.S95E5 2000). BC.: Ministry of
Environment, Land, and Parks.
British Petroleum (BP). (2012). Conversion Factors. Retrieved March 12, 2012, from
http://www.bp.com/conversionfactors.jsp.
Bureau of Ocean and Energy Management (BOEM). (undated). Oil-­‐Spill Occurrence Rate for
Oil-­‐Spill Risk Analysis (OSRA). Retrieved from
http://www.boem.gov/Environmental-­‐Stewardship/Environmental-­‐
Assessment/Oil-­‐Spill-­‐Modeling/Oil-­‐Spill-­‐Occurence-­‐Rate-­‐for-­‐Oil-­‐Spill-­‐Risk-­‐Analysis-­‐
(OSRA).aspx.
Canadian Environmental Assessment Act (CEAA), S.C. 2012, c. 19, s. 52.
Caza, A. (2012). Personal Communication with S. Broadbent.
Commissioner of the Environment and Sustainable Development (CESD). (2011). Report of
the Commissioner of the Environment and Sustainable Development. ( No. FA1-­‐
2/2011-­‐2-­‐0E-­‐PDF). Ottawa, ON.: Office of the Auditor General of Canada.
Det Norske Veritas (DNV). (2004). Ports and the ISPS Code -­‐ will they make it? Retrieved
from
http://www.dnv.com/industry/maritime/publicationsanddownloads/publications/
maritime_magazine/2004/class_news_1_2004/PortsandtheISPSCodewilltheymakeit
.asp.
Det Norske Veritas (DNV). (2008). It Takes Management Commitment to Reduce Accidents.
Retrieved from
http://www.dnv.com/industry/maritime/publicationsanddownloads/publications/
dnvtankerupdate/2008/no12008/ittakesmanagementcommitmenttoreduceacciden
ts.asp.
Dudding, P. (2013). Personal Communication with S. Broadbent.
Dynamic Risk Assessment Systems, Inc. (DRAS). (2012). Threat Assessment. In
WorleyParsons Canada Services Limited (WorleyParsons) (Ed.), Semi-­‐Quantitative
Risk Assessment (pp. 142).
Enbridge. (2001). 2001 Environment, Health and Safety Report. Retrieved 2012, June 14,
from
http://www.enbridge.com/AboutEnbridge/CorporateSocialResponsibility/~/medi
a/www/Site%20Documents/About%20Enbridge/Corporate%20Social%20Respon
sibility/Reports/2001-­‐ehs-­‐report.ashx.
78

Enbridge. (2004). 2004 Corporate Social Responsibility Report. Retrieved 2012, June 14,
from
http://www.enbridge.com/AboutEnbridge/CorporateSocialResponsibility/~/medi
a/www/Site%20Documents/About%20Enbridge/Corporate%20Social%20Respon
sibility/Reports/2004-­‐csr.ashx.
Enbridge. (2007). 2007 Corporate Social Responsibility Report. Retrieved 2012, June 14,
from
http://www.enbridge.com/AboutEnbridge/CorporateSocialResponsibility/~/medi
a/www/Site%20Documents/About%20Enbridge/Corporate%20Social%20Respon
sibility/Reports/enbridgecsr2007.ashx.
Enbridge. (2010a). 2010 Corporate Social Responsibility Report. Retrieved 2012, July 15,
from http://csr.enbridge.com/csr2010/downloads/2010_Enbridge_CSR_Report.pdf.
Enbridge. (2010b). Enbridge Northern Gateway Project Section 52 Application, Volumes 1-­‐
8. Retrieved from http://gatewaypanel.review-­‐examen.gc.ca/clf-­‐
nsi/pblcrgstr/pblcrgstr-­‐eng.html.
Enbridge. (2010c). TERMPOL Study No. 3.15: General Risk Analysis and Intended Methods
of Reducing Risk. Retrieved from http://gatewaypanel.review-­‐examen.gc.ca/clf-­‐
nsi/pblcrgstr/pblcrgstr-­‐eng.html.
Enbridge. (2011). 2011 Corporate Social Responsibility Report. Retrieved July 15, 2012,
from http://csr.enbridge.com/downloads/pdf/Enbridge-­‐2011-­‐CSR-­‐Report.pdf.
Enbridge. (2012a). Northern Gateway Response to Coastal First Nations’ Request for
Follow-­‐Up Information on Reply Evidence. Retrieved from
http://gatewaypanel.review-­‐examen.gc.ca/clf-­‐nsi/pblcrgstr/pblcrgstr-­‐eng.html.
Enbridge. (2012b). ENGP Hearing Transcripts. Retrieved from
http://gatewaypanel.review-­‐examen.gc.ca/clf-­‐nsi/pblcrgstr/pblcrgstr-­‐eng.html.
Environment Canada Oil Properties Database (EC OPD). (2012). Oil Properties for
Athabasca Bitumen and West Texas Intermediate. Retrieved 2012, March 6, from
http://www.etc-­‐cte.ec.gc.ca/databases/OilProperties.
Eliopoulou, E., Papanikolaou, A. (2007). Casualty Analysis of Large Tankers, Journal of
Marine Science and Technology, Vol. 12, 240-­‐250.
Eliopoulou, E., Papanikolaou, A., Diamantis, P., Hamann, R. (2011). Analysis of Tanker
Accidents after OPA 90, acc. for publication at the Journal of Engineering for the
Maritime Environment (JEME).
79

Eschenbach, T. G., Harper, W. V., Anderson, C. M., & Prentki, R. (2010). Estimating Oil Spill
Occurrence Rates: A Case Study for Outer Continental Shelf Areas of Gulf of Mexico.
Journal of Environmental Statistics, 1(1).
FORCE Technology (FORCE). (2010). Manoeuvring Study of Escorted Tankers to and from
Kitimat: Real-­‐time Simulations of Escorted Tankers bound for a Terminal at Kitimat.
FORCE Technology no. 108-­‐29930 ES, Version 4.0. Prepared for Northern Gateway
Pipelines LP. Retrieved from http://gatewaypanel.review-­‐
examen.gc.ca.proxy.lib.sfu.ca/clf-­‐nsi/pblcrgstr/pblcrgstr-­‐eng.html.
Glosten. (2004). Study of Tug Escorts in Puget Sound. Prepared for State of Washington:
Department of Ecology Lacey, Washington. File No. 04075. The Glosten Associates:
Seattle, WA.
Gunton, T. I., & Broadbent, S. C. (2012a). A Public Interest Assessment of the Enbridge
Northern Gateway Project. Report Prepared for Coastal First Nations.
Gunton, T. I., & Broadbent, S. C. (2012b). A Review of Potential Impacts to Coastal First
Nations from an Oil Tanker Spill Associated with the Northern Gateway Project.
Report Prepared for Coastal First Nations.
Hassel, M., Asbjornslett, B. E., & Hole, L. P. (2011). Underreporting of Maritime Accidents to
Vessel Accident Databases. Accident Analysis and Prevention, 43(6), 2053-­‐2063.
Hoekstra, G. (2012, January 3). First Nations Fiercely Opposed to Northern Gateway. The
Vancouver Sun, Retrieved from
http://www.vancouversun.com/news/First+nations+fiercely+opposed+Northern+
Gateway/5937416/story.html.
International Association of Marine Aids to Navigation and Lighthouse Authorities (IALA).
(2008). IALA Guideline 1018 on Risk Management. Retrieved 2012, March 2, from
http://www.iala-­‐aism.org/iala/publications/documentspdf/doc_92_eng.pdf.
LaBelle, R. P. (2012). Personal communication with S. Broadbent.
Living Oceans. (2011). Northern Gateway Project Tanker Routes. Retrieved 2013, March
10, from http://www.livingoceans.org.
Lyons, W. C., & Plisga, G. S. (2005). Standard Handbook of Petroleum & Natural Gas
Engineering Elsevier/Gulf Professional Pub. Retrieved from
http://search.ebscohost.com/login.aspx?direct=true&db=nlebk&AN=117161&site=
ehost-­‐live.
80

MacDonald, K. (2012). Northern Gateway Pipelines Application to the National Energy
Board Enbridge Northern Gateway Project OH-­‐4-­‐2011, NEB File No: OF-­‐Fac-­‐Oil-­‐
N304-­‐2010-­‐01 01, JRP Information Request (“IR”) 8.1. Retrieved 2012, June 12, from
http://gatewaypanel.review-­‐examen.gc.ca/clf-­‐nsi/pblcrgstr/pblcrgstr-­‐eng.html.
National Energy Board (NEB). (2011). Focus on Safety and Environment: A Comparative
Analysis of Pipeline Performance, 2000‐2009. Retrieved from
http://publications.gc.ca/collections/collection_2012/one-­‐neb/NE2-­‐2-­‐2009-­‐
eng.pdf.
National Research Council (NRC). (1996). Understanding Risk: Informing Decisions in a
Democratic Society. Washington, DC.: National Academies Press.
National Transportation Safety Board (NTSB). (1990). Marine Accident Report: Grounding
of the US Tankship Exxon Valdez: on Bligh Reef, Prince William Sound, near Valdez,
Alaska, March 24, 1989. (No. NTSB/MAR-­‐90/04). Washington, D.C.: NTSB.
National Transportation Safety Board (NTSB). (2012). Emergency and Environmental
Response Group Chairman’s Factual Report. (No. NTSB No.: DCA-­‐10-­‐MP-­‐007).
Washington, D.C.: National Transportation Safety Board.
Price, J. M., Johnson, W. R., Marshall, C. F., Ji, Z., & Rainey, G. B. (2003). Overview of the Oil
Spill Risk Analysis (OSRA) Model for Environmental Impact Assessment. Spill
Science & Technology Bulletin, 8(5-­‐6), 529-­‐533.
Papanikolaou, A., Eliopoulou, E., Hamann, R., Loer, K. (2009). Assessment of Safety of Crude
Oil Transport by Tankers, Proc. Annual Main Conference of Schiffbautechnische
Gesellschaft (STG2009), Berlin, November.
Psarros, G., Skjong, R., & Eide, M. S. (2010). Under-­‐reporting of Maritime Accidents.
Accident Analysis and Prevention, 42(2), 619.
Sea Distances -­‐ Voyage Calculator. (undated). Retrieved from http://sea-­‐distances.com.
Smith, R. A., Slack, J. R., Wyant, T. & Lanfear, K. J. (1982). The Oilspill Risk Analysis Model of
the U.S. Geological Survey. Geological Survey Professional Paper 1227. Retrieved
from http://www.boemre.gov/eppd/PDF/smithetal.pdf.
Transport Canada (TC). (2010). Port State Control: 2010 Annual Report. Retrieved 2013,
March 20, from
http://www.tc.gc.ca/media/documents/marinesafety/TP13595E.pdf.
Transport Canada (TC). (2013). Harper government announces first steps towards World-­‐
Class Tanker Safety System. Retrieved 2013, March 19, from
http://www.tc.gc.ca/eng/mediaroom/releases-­‐2013-­‐h031e-­‐7089.htm.
81

United States Coast Guard, National Research Center (USCG NRC). (2012). Reported Spills
for the Keystone Pipeline from 2010 -­‐ 2011 (Data file). Retrieved from
http://www.nrc.uscg.mil/pls/apex/f?p=109:2:0::NO:::.
United States Department of Interior, Minerals Management Service (US DOI). (1997).
Revised Oil-­‐Spill Risk Analysis: Beaufort Sea Outer Continental Shelf Lease Sale 170.
MMS-­‐97-­‐0039. Retrieved from www.boemre.gov/itd/pubs/1997/97-­‐0039.pdf.
United States Department of Interior, Minerals Management Service (US DOI). (1998). Oil-­‐
Spill Risk Analysis: Destin Dome Development and Production Plan. MMS-­‐98-­‐0037.
Retrieved from http://www.boemre.gov/itd/pubs/1998/98-­‐0037.pdf.
United States Department of Interior, Minerals Management Service (US DOI). (2000a). Oil-­‐
Spill Risk Analysis: Pacific Outer Continental Shelf Program. MMS-­‐2000-­‐057.
Retrieved from www.boemre.gov/itd/pubs/2000/2000-­‐057.pdf.
United States Department of Interior, Minerals Management Service (US DOI). (2000b). Oil-­‐
Spill Risk Analysis: Liberty Development and Production Plan. MMS-­‐2000-­‐082.
Retrieved from www.boemre.gov/itd/pubs/2000/2000-­‐082.pdf.
United States Department of Interior, Minerals Management Service (US DOI). (2002). Oil-­‐
Spill Risk Analysis: Gulf of Mexico Outer Continental Shelf (OCS) Lease Sales, Central
Planning Area and Western Planning Area, 2003-­‐2007, and Gulfwide OCS Program,
2003-­‐2042. MMS-­‐2002-­‐032. Retrieved from
www.boemre.gov/itd/pubs/2002/2002-­‐032.pdf.
United States Department of Interior. Minerals Management Service (US DOI). (2003a).
Cook Inlet Planning Area. Oil and Gas Lease Sales 191 and 199. Final Environmental
Impact Statement. MMS-­‐2003-­‐055. Anchorage, Alaska.
United States Department of Interior, Minerals Management Service (US DOI). (2003b). Oil-­‐
Spill Risk Analysis: Cook Inlet Planning Area, OCS Lease Sales 191 and 199. MMS
2002-­‐074. Retrieved from www.boemre.gov/itd/pubs/2002/2002-­‐074.pdf.
United States Department of Interior, Minerals Management Service (US DOI). (2007a). Oil-­‐
Spill Risk Analysis: Sivulliq Exploration Project. MMS-­‐2007-­‐039. Retrieved from
www.boemre.gov/itd/pubs/2007/2007-­‐039.pdf.
United States Department of Interior, Minerals Management Service (US DOI). (2007b). Oil-­‐
Spill Risk Analysis: Gulf of Mexico Outer Continental Shelf (OCS) Lease Sales, Central
Planning Area and Western Planning Area, 2007-­‐2012, and Gulfwide OCS Program,
2007-­‐2046. MMS-­‐2007-­‐040. Retrieved from
www.boemre.gov/itd/pubs/2007/2007-­‐040.pdf.
82

United States Department of Transportation, Pipeline and Hazardous Materials Safety
Administration (US PHMSA). (2011). Reporting Criteria Changes -­‐ 1990-­‐Current.
Retrieved 2012, June 15, from
http://primis.phmsa.dot.gov/comm/reports/safety/docs/IncidentReportingCriteri
aHistory1990-­‐2011.pdf.
United States Department of Transportation, Pipeline and Hazardous Materials Safety
Administration (US PHMSA). (2012). PHMSA Hazardous Liquid Flagged Incidents
File. Retrieved 2012, March 2, from
http://primis.phmsa.dot.gov/comm/reports/safety/SIDA.html?nocache=5779.
United States Department of State (USDS). (2011). Final Environmental Impact Statement
for the Keystone XL Project. Washington, DC.: United States Department of State.
United States Environmental Protection Agency (US EPA). (1998). Guidelines for Ecological
Risk Assessment. (No. EPA/630/R-­‐95/002F). Washington, DC.: United States
Environmental Protection Agency.
United States Environmental Protection Agency (US EPA). (2000). Risk Characterization
Handbook. (No. 100-­‐B-­‐00-­‐002). Washington, DC.: United States Environmental
Protection Agency.
United States Environmental Protection Agency (US EPA). (2004). Risk Assessment
Principles & Practices. (Staff Paper Prepared for the U.S. Environmental Protection
Agency by members of the Risk Assessment Task Force No. 100/B-­‐04/001).
Washington, DC.: United States Environmental Protection Agency.
WorleyParsons Canada Services Limited (WorleyParsons). (2012). Semi-­‐Quantitative Risk
Assessment. Retrieved 2012, June 12, from http://gatewaypanel.review-­‐
examen.gc.ca/clf-­‐nsi/pblcrgstr/pblcrgstr-­‐eng.html.
Wright Mansell (WM). (2012). Public Interest Benefit Evaluation of the Enbridge Northern
Gateway Pipeline Project: Update and Reply Evidence. Retrieved from
http://gatewaypanel.review-­‐examen.gc.ca/clf-­‐nsi/pblcrgstr/pblcrgstr-­‐eng.html.
83

Appendix A -­‐ Detailed Summary of ENGP Spill Risk Analysis
This appendix provides a detailed summary of the methodologies used by Enbridge and its
consultants to estimate spill likelihood for ENGP tanker, terminal, and pipeline spills. This
summary relies on the same regulatory documents referenced in section 3 in the main
body of the report.
1. Spills from Tanker Traffic Accessing Kitimat Terminal
DNV determines spill return periods for tanker traffic accessing Kitimat Terminal in the
Marine Shipping Quantitative Risk Analysis prepared for the Northern Gateway Project
on behalf of Enbridge. DNV uses a per-­‐voyage methodology to forecast spills in the
marine shipping QRA based on nautical miles (nm) travelled by tankers within the
study area. The methodology consists of the following steps:
• Determine base incident frequencies per nm
• Apply scaling factors to base incident frequencies to adjust for local conditions
• Determine conditional spill probabilities from an assessment of incident
consequences
• Calculate unmitigated spill return periods based on scaled incident frequencies
and conditional probabilities
• Conduct sensitivity analysis
• Identify mitigation measures and recalculate spill return periods based on
mitigation
Base Incident Frequencies
As part of the per-­‐voyage methodology, DNV determines base incident frequencies from
historical tanker incident data and uses various assumptions to estimate incident
frequencies per nm. DNV determines frequencies for international tanker incidents
based on data from Lloyds Register Fairplay (LRFP) for the period 1990-­‐2006
(Brandsæter and Hoffman 2010 p. 5-­‐49). DNV examines four types of potential
incidents associated with tanker traffic:
1. Powered and drift grounding, whereby a tanker runs aground either with
functional mechanical and navigation equipment (powered) or as a result of
failed propulsion equipment (drift)
2. Collisions, whereby the navigational failure of one or both vessels results in a
collision
3. Foundering, whereby a tanker sinks due to structural failure of the hull from
weather or structural defects
4. Fire and/or explosion.
As shown in Table A-­‐1, DNV calculates LRFP incident frequencies per nm based on the
following assumptions of the average distance travelled by a tanker per year for each
tanker incident type:
84

• Tankers sail in coastal areas where a powered grounding may occur 10% of the
time at sea (Rabaska 2004 as cited in Brandsæter and Hoffman 2010)
• Tankers sail in areas with heavy traffic where collisions may occur 20% of the
time at sea (Rabaska 2004 as cited in Brandsæter and Hoffman 2010)
• Tankers sail in areas where land is within drifting distance and drift grounding
may occur if the ship loses power 15% of the time at sea
• Tankers sail in open water where foundering may occur 90% of the time at sea.
DNV supports the first two assumptions with reference to a TERMPOL Study completed
for Rabaska (2004), while the assumptions for drift grounding and foundering are not
supported with evidence.
Table A-­‐1: Base Incident Frequencies for Tanker Incidents based on Per-­‐Voyage Methodology
Powered
Grounding
Drift
Grounding
Base LRFP incident frequency
(per ship year*)
Average distance sailed by
74,022
tanker (nm)
Portion of total distance sailed
by where incident may occur
Distance sailed by tanker
where incident may occur
Base LRFP incident frequency
(per nm)
Source: Brandsæter and Hoffman (2010 p. 5-­‐51).
* One ship year represents a total sailing distance of 74,022 nm per year per tanker in operation.
Collision Foundering
Fire/
Explosion
4.42E-­‐03 1.11E-­‐03 6.72E-­‐03 3.36E-­‐05 2.41E-­‐03
10% 15% 20% 90% 100%
7,402 11,103 14,804 66,620 74,022
5.98E-­‐07 9.96E-­‐08 4.54E-­‐07 5.04E-­‐10 3.26E-­‐08
Local Scaling Factors
DNV then calculates incident data for the BC study area by multiplying the international
frequency data from LRFP by scaling factors that compare risks in the study area to the
international areas on which the incident data are based. DNV identified scaling factors
from a qualitative assessment of the international data and the scaling factors
underwent a review by experts familiar with marine risk assessment, tanker
operations, and global navigation (Brandsæter and Hoffman 2010 p. 5-­‐52). DNV uses
the following formula to calculate incident frequencies for the BC study area where
FrequencyBC-­‐Coast is the scaled incident frequency per nm, Fbase is the base incident
frequency per nm from LRFP in Table A-­‐1, and Klocal-­‐scaling-­‐factor is the total of local scaling
factors identified in Table A-­‐2:
Frequency BC−Coast = F base * K local−scaling− factor
DNV divides the BC study area into nine segments and assesses scaling factors for each
incident type. The comparison uses 1.0 as a baseline and thus a scaling factor equal to
1.0 suggests that the local conditions affecting risk are equal to the world average
(Brandsæter and Hoffman 2010 p. 5-­‐52). Risk factors assessed by DNV depend on the
type of incident and include the following categories:
85

• Navigational route: number of course changes
• Measures: use of pilots with local knowledge
• Navigational difficulty: visibility, currents, and disturbance from other vessels
• Distance to shore: distance from ship to shore
• Emergency anchoring: possibility to drop an emergency anchor
• Traffic density: Number of vessels encountered during transit
• Weather conditions: Accounts for winds and currents.
Table A-­‐2 summarizes scaling factors used by DNV for each incident type. DNV applied
three scaling factor categories to assess powered groundings and collisions, two scaling
factors to assess drift groundings, and one factor (weather) to assess foundering
incidents. Scaling factors range from 0.001 to 1.94 for all incident types. The overall
average scaling factor for each incident type ranges between 0.22 for collision incidents
and 1.11 for drift grounding incidents, with the exception of incidents involving fire or
explosion, which DNV does not scale to represent local conditions. DNV does not
provide any information on the routes that the ENGP routes are being compared to so it
is not possible to assess the basis or accuracy of the scaling factors.
Table A-­‐2: Local Scaling Factors for Tanker Incident Frequencies
Incident
Type
Powered
Grounding
Drift
Grounding
Navigational
Route
Distance to
Shore
Scaling Factor Category
Measures
Emergency
Anchoring
Collision Traffic Density Measures
Foundering
Weather
Conditions
Fire/
Explosion
Source: Brandsæter and Hoffman (2010).
* Calculated based on data from Brandsæter and Hoffman (2010).
Navigational
Difficulty
Range of Total
Scaling Factors
Average
Scaling Factor*
0.001 – 1.94 0.91
-­‐ 0.01 – 1.56 1.11
Navigational
Difficulty
0.01 – 0.54 0.22
-­‐ -­‐ 0.01 – 1.50 0.80
Conditional Probabilities
After estimating the frequency of a tanker incident occurring by multiplying scaling
factors times the international incident frequency rate, DNV determines the probability
of a spill resulting from a tanker incident. DNV uses two different methods to estimate
conditional spill probabilities.
Method 1
The first method (Method 1) determines conditional spill probabilities based on LFRP
data. According to Brandsæter and Hoffman (2010), LRFP categorize damages for the
four incident types based on the following categories:
• Minor Damage: Damages cause small indents to the vessel but do not penetrate
the outer hull; there is very minor damage involving little repair
n/a
86

• Major Damage: Damage to the tanker penetrates the outer hull but the spill
event is not categorized as total loss; structural, mechanical, or electrical damage
that require repairs; a breakdown results in the tanker being towed
• Total Loss: The tanker ceases to exist after an accident; cost of repairs exceeds
insured value of ship; all cargo is lost.
LRFP data provides a frequency distribution for each damage category identified above
depending on the incident type (Table A-­‐3). For example, the frequency distribution of
damages for a grounding incident suggests that 57.2% of incidents result in minor
damage, 40.4% result in major damage, and only 2.4% of incidents result in total loss of
the tanker (Brandsæter and Hoffman 2010 p. 6-­‐79).
DNV determines conditional spill probabilities for each incident type and damage
category. According to DNV, “The conditional probability of a spill has been estimated
by DNV based on the research of spill to damage data” (Brandsæter and Hoffman 2010
6-­‐79). DNV then multiplies frequency distributions by probabilities of a spill for laden
tankers and tankers in ballast for each damage category to determine the total
conditional spill probabilities for each incident type. DNV disaggregates conditional
probabilities presented in Table A-­‐3 by laden tankers carrying oil or condensate and
tankers in ballast carrying bunker fuel. For example, the conditional spill probability
for a laden tanker running aground is (57.2% x 0%) + (40.4% x 75%) + (2.4% x 100%)
= 32.7% and thus a spill is predicted 32.7% of the time a grounding incident occurs that
involves a laden tanker.
Table A-­‐3: Damage Frequencies and Conditional Probabilities for Tanker Incidents (Method 1)
Incident Damage LRFP Frequency Conditional Probability of Spill (%)
Type Category Distribution (%)
Laden Ballast
Minor 57.2 0 0
Grounding
Major
Total Loss
40.4
2.4
75
100
10
100
Total Conditional Probability 32.7 6.4
Minor 74.5 0 0
Collision
Major
Total Loss
25.5
Negligible
75
100
10
100
Total Conditional Probability 19.1 2.6
Minor 0 -­‐ -­‐
Foundering
Major
Total Loss
0
100
-­‐
100
-­‐
100
Total Conditional Probability 100 100
Minor 48.8 0 0
Fire/ Major 48.4 50 10
Explosion Total Loss 2.8 100 100
Total Conditional Probability
Source: Brandsæter and Hoffman (2010 p. 6-­‐79; 6-­‐83; 6-­‐87; 6-­‐88).
27.0 7.6
Method 2
The second method (Method 2) estimates spill quantities for bottom and side damages
based on information from the International Marine Organization International
87

Convention for the Prevention of Pollution from Ships (MARPOL) and estimates spill
size distribution with Naval Architecture Package software. Furthermore, Method 2
uses Monte Carlo simulation to estimate probabilities for various spill sizes for
groundings and collisions.
Table A-­‐4 presents results for probabilities and spill volumes calculated with Method 2.
According to DNV calculations, if the largest laden tanker (VLCCmax) forecast to call at
Kitimat Terminal ran aground, there is an 18.7% probability that the damaged vessel
would spill oil. DNV also estimates that the average amount of oil released from a
VLCCmax spill due to grounding is 1,725 m 3 , while an extreme outflow -­‐ defined as the
mean of the highest outflow decile -­‐ is 15,506 m 3 (Brandsæter and Hoffman 2010 p. 6-­‐
80).
Table A-­‐4: Conditional Spill Probabilities and Spill Volumes for Grounding and Collision Incidents
(Method 2)
Tanker Type
Laden
Vessel
Vessel
in
Ballast
Conditional
Spill
Probability
(%)
Grounding Collision
Mean
Oil
Outflow
(m3) Extreme
Oil
Outflow
(m3) Conditional
Spill
Probability
(%)
Mean
Oil
Outflow
(m3) Extreme
Oil
Outflow
VLCCmax 18.7 1,725 15,506 23.9 1,399 35,839
VLCC 18.7 1,616 14,469 24.9 1,397 35,605
Suezmax 17.5 1,106 9,481 27.7 1,280 28,980
Aframax 18.0 736 6,710 21.0 638 17,539
VLCC 0.2 1.01 11 8.3 210 2,101
Suezmax < 0.1 0.01 0.11 5.7 86 860
Aframax < 0.1 0.08 0.7 14.2 174 1,414
Source: Brandsæter and Hoffman (2010 p. 6-­‐80; 6-­‐84).
Note: Extreme oil outflow represents the mean of the 10% of bottom damage incident with the most outflow.
For grounding and collision incidents, Method 1 and Method 2 estimate two sets of
conditional probabilities. Spill probabilities of 17.5% to 18.7% for laden tankers
running aground compare with the conditional probability of 32.7% for the same
incident type determined in Method 1 using LRFP data. DNV states that, due to the
rocky local seabed conditions, the higher conditional probability of 32.7% is used in
subsequent chapters to determine unmitigated and mitigated spill probabilities
(Brandsæter and Hoffman 2010 p. 6-­‐81). Similarly, the probability of a spill from a
collision for a laden tanker in Method 1 is 19.1% and conditional probabilities for laden
vessels calculated using Method 2 range between 21.0% and 27.7%. DNV uses the
lower probability of 19.1% claiming that, compared to the world average, the
probability of a tanker being struck at a perpendicular angle in the cargo or fuel tank
area of the vessel is lower due to the local traffic pattern (Brandsæter and Hoffman
2010 p. 6-­‐84).
In addition to determining spill probabilities for various tanker sizes and mean and
extreme oil outflows, Method 2 also estimates probabilities for grounding and collision
spills exceeding a certain volume. According to the Monte Carlo simulation, the
probability of a spill greater than 10,000 m 3 for a laden VLCC grounding is 5.5% and
(m 3)
88

25% for a collision (Brandsæter and Hoffman 2010 p. 6-­‐81 -­‐ 6-­‐85). The probability of a
grounding spill exceeding 25,000 m 3 and 40,000 m 3 is 1% and 0.2%, respectively, while
the probability of a collision spill exceeding 20,000 m 3 and 50,000 m 3 is 10% and 0.3%,
respectively (Brandsæter and Hoffman 2010 p. 6-­‐81 -­‐ 6-­‐85).
Unmitigated Spill Return Periods
DNV estimates unmitigated spill return periods for each type of tanker incident based
on scaled incident frequencies, conditional probabilities, the length of each segment,
and the number of times the route is travelled per year. DNV presents unmitigated risk
as a return period, which is the number of years between spill events, and is an
alternative to stating the annual probability of a spill. The following equation
represents the return period of a spill for a particular segment (ReturnPeriodsegment i)
where Fi,j is the frequency of incident type j in segment i, di,j is the conditional
probability of a spill given incident type j in segment i, Xi is the sailing distance in nm
through segment i, and n is the annual number of times the route through segment i is
travelled:
ReturnPeriod segmenti =1/[Σ(F i, j * d i, j )* X i * n]
DNV calculates return periods for unmitigated spills based on forecast traffic of 149 oil
tankers and 71 condensate tankers and the distribution of tanker traffic on the
northern and southern routes identified by Enbridge (see Table A-­‐5). According to DNV,
vessels will have the option of selecting which shipping route to access Kitimat
Terminal and will select the route based on weather conditions and their final
destination (Brandsæter and Hoffman 2010 p. 3-­‐10). However, DNV allocates the
majority of traffic (118 of 220 tankers or 54%) in its analysis to Caamano Sound, which
has the shortest route distance of 190 nm and hence a lower return period than longer
routes.
Table A-­‐5: Distribution of Tanker Traffic Per Year
Route and Distance (nm) Oil Condensate Total
North Route (251 nm) 73 0 73
South Route via Caamano Sound (190 nm) 61 57 118
South Route via Browning Entrance (259 nm) 15 14 29
Total 149 71 220
Source: Brandsæter and Hoffman (2010 p. 3-­‐12; p. 7-­‐105).
Table A-­‐6 presents unmitigated spill return periods for each type of tanker incident.
According to DNV, an unmitigated tanker incident will result in an oil/condensate spill
once every 78 years and the return period for an unmitigated oil spill is 110 years.
89

Table A-­‐6: Unmitigated Return Periods for an Incident Resulting in a Spill
Incident Type
Total Return Period (in years)
Oil/Condensate Spill Oil Spill
Powered Grounding* 107 154
Drift Grounding* 535 742
Collision* 1,084 1,546
Foundering* 25,821 37,963
Fire/ Explosion* 1,838 2,623
Total
Source: Brandsæter and Hoffman (2010).
78 110
* Calculated from Brandsæter and Hoffman (2010).
Note: Total return period represents forecast traffic on all three routes.
DNV also estimates unmitigated return periods for various oil spill sizes. Based on oil
spill size calculations for grounding and collisions estimated in Method 2 of the
consequence assessment, DNV expects that over 41% of spills will exceed 5,000 m 3 ,
while nearly 5% and 1% of spills will exceed 20,000 m 3 and 40,000 m 3 , respectively
(Table A-­‐7). DNV determines that the unmitigated return period for oil spills greater
than 5,000 m 3 is 200 years, while return periods for oil spills greater than 20,000 m 3
and 40,000 m 3 are 1,750 years and 12,000 years, respectively. DNV does not provide
any rationale for the spill volume categories it uses to estimate these return periods and
does not provide estimates for return periods for specific spill volume categories less
than 5,000 m 3 , even though spills as small as 238 m 3 can have significant adverse
environmental impacts (US DOI 2003a).
Table A-­‐7: Unmitigated Oil Spill Size Distributions and Return Periods
Spill Volume (m3) Unmitigated
Distribution (%)
Estimated Return
Period (in years)
> 5,000 41.5 200
> 20,000 4.5 1,750
> 40,000 0.8 12,000
Source: Calculated from Brandsæter and Hoffman (2010 p. 7-­‐110).
Note: Estimates for unmitigated distribution of spill sizes should be considered as approximate estimates due to limited data
provided by DNV.
Sensitivity Analysis for Unmitigated Spill Return Periods
DNV conducts a sensitivity analysis in its marine shipping QRA examining impacts of
several parameters on spill probabilities. The four parameters include: (1) increased
scaling factor for grounding, (2) increased traffic density, (3) increased/decreased
number of tankers calling Kitimat Terminal, and (4) extending segments 5 and 8 to 200
nm. The sensitivity analysis tests the impact of changes in parameters on all three
potential routes separately under the assumption that all 220 tankers forecast to call at
Kitimat Terminal use the route being tested exclusively. Therefore, the spill
probabilities do not account for using a combination of routes.
Sensitivity 1: Increased Scaling Factors for Grounding Incidents
The first parameter in the sensitivity analysis evaluates the effect of increasing scaling
factors for powered and drift grounding by 20% for each of the nine route segments on
90

the overall tanker spill return periods. DNV evaluates the sensitivity of this particular
parameter because grounding is the most significant hazard that may occur on
segments that have a high contribution to the overall risk per route (Brandsæter and
Hoffman 2010 p 7-­‐99). DNV provides no rationale for using the 20% increase in local
scaling factors for grounding.
Table A-­‐8 presents a comparison of the unmitigated return periods for baseline
conditions and increased scaling factors for grounding for each tanker route. Return
periods represent the 20% increase in scaling factors for grounding while leaving other
scaling factors for other tanker incident types including collision, foundering, and fire
and/or explosion unchanged. According to DNV calculations, increased scaling factors
for powered and drift grounding incidents decrease return periods as much as 13 years
for the South Route via Browning Entrance.
Table A-­‐8: Sensitivity Analysis for a 20% Increase in Scaling Factors for Tanker Grounding
Incidents
Sensitivity Parameter North Route
South Route via
Caamano Sound
South Route via
Browning
Entrance
Return Periods with Baseline Scaling
Factors (in years)
69 83 84
Return Periods with 20% Increase in
Scaling Factors for Grounding (in years)
Source: Brandsæter and Hoffman (2010 p. 7-­‐100).
59 71 71
Note: Return periods represent all incident types for each route segment.
Sensitivity 2: Increased Traffic that Affects Collision Frequency
DNV examines the effect of increased traffic that affects collision frequency along the
three routes on overall spill return periods. Scaling factors for traffic density for
collision incidents increase between 0% and 50% for each route segment that,
according to DNV, allows for the increase in forecast developments in Kitimat and the
general increase in coastal vessel traffic (Brandsæter and Hoffman 2010 p. 7-­‐100).
Return periods for increased traffic densities for collision incidents also represent
baseline scaling factors for powered and drift grounding, foundering, and fire and/or
explosion. Based on the higher traffic density for collision incident frequencies, the
overall return period for each tanker route decreases marginally by two years (Table A-­‐
9).
According to DNV, traffic density is “…relatively low, even with further projects taken
into consideration” (Brandsæter and Hoffman (2010 p. 7-­‐100). However, DNV does not
provide a summary of future projects that it took into consideration and provides no
evidence how it to relates modified scaling factors to potential increases in traffic at or
near the Port of Kitimat referenced in the TERMPOL Study 3.2: Origin, Destination, &
Marine Traffic Volume Survey. Despite increases, scaling factors for traffic density all
remain well below the international average of 1.0. Furthermore, DNV does not clearly
state whether the base case scaling factors that represent traffic density include
planned tanker traffic associated with the ENGP.
91

Table A-­‐9: Sensitivity Analysis for Increased Traffic Density for Tanker Collision Incidents
Sensitivity Parameter North Route
South Route via
Caamano Sound
South Route via
Browning
Entrance
Return Periods with Baseline Scaling
Factors (in years)
69 83 84
Return Periods with Increased Traffic
Density for Collisions (in years)
Source: Brandsæter and Hoffman (2010 p. 7-­‐101).
67 81 82
Note: Return periods represent all incident types for each route segment.
Sensitivity 3: Increased/Decreased Number of Tankers
A sensitivity analysis examines the impact of a decrease in annual average tanker traffic
from 220 vessels to 190 tankers and an increase to 250 tankers per year. DNV fails to
provide a rationale for using this particular range for the number of vessels in its
sensitivity analysis. It is important to note that the higher scenario of 250 tankers is
still well below the planned expansion capacity to 1,125 kbpd that requires 331
tankers. As shown in Table A-­‐10, increasing the number of sailings from 220 tankers to
250 reduces the return period by 8 to 10 years depending on the route (Brandsæter
and Hoffman 2010 p. 7-­‐102).
Table A-­‐10: Sensitivity Analysis for Changes in the Number of Tankers Calling Kitimat Terminal
Sensitivity Parameter North Route
South Route via
Caamano Sound
South Route via
Browning Entrance
190 Tankers 80 96 97
220 Tankers (Baseline) 69 83 84
250 Tankers 61 73 74
Source: Brandsæter and Hoffman (2010 p. 7-­‐102).
Sensitivity 4: Route Extension to 200 Nautical Miles
DNV’s final sensitivity analysis examines increases in the distance sailed by tankers in
two segments of the northern and southern routes. DNV assumes that tankers
navigating segment 5 and 8 travel a distance of 200 nm to reach the open Pacific Ocean
as opposed to the baseline distance of 65 nm and 75 nm for each segment. DNV further
assumes that incident frequencies and scaling factors for each incident type remain
unchanged for the route extension to 200 nm in segments 5 and 8, even though DNV
acknowledges that tankers travelling the extended route will navigate through areas
with greater vessel traffic (Brandsæter and Hoffman 2010 p. 7-­‐103).
Table A-­‐11 compares spill return periods for an extension to 200 nm in segments 5 and 8
with baseline distances. An increase in segments 5 and 8 to 200 nm reduces the return
period to 1,400 years and 1,500 years, respectively, for those particular segments. DNV
does not calculate the effect of increased sailing distances in segment 5 and 8 on the
overall unmitigated return period for each route.
92

Table A-­‐11: Sensitivity Analysis for Route Extension on Segments 5 and 8
Segment 5 Segment 8
Sensitivity Return Period (in years) Return Period (in years)
Parameter Baseline Extension Baseline Extension
65 nm
200 nm
75 nm
200 nm
Route
Extension
4,200 1,400 4,100 1,500
Source: Brandsæter and Hoffman (2010 p. 7-­‐103).
In summary, DNV’s sensitivity analysis examines the effects of changes to four inputs on
spill return periods. Results from the sensitivity analysis suggest that return periods
change modestly compared to baseline return periods (Table A-­‐12).
Table A-­‐12: Summary of Sensitivity Analysis on Spill Return Periods (years)
Sensitivity Parameter North Route
South Route
via
Caamano
Sound
South Route
via
Browning
Entrance
Baseline 69 83 84
Sensitivity 1: Increase in Scaling Factors for Grounding 59 71 71
Sensitivity 2: Increase in Traffic Density for Collisions 67 81 82
Sensitivity 3a: Decrease to 190 Tankers per year 80 96 97
Sensitivity 3b: Increase to 250 Tankers per year 61 73 74
Sensitivity 4: 200 nm Extension to Segments 5 and 8* 67 81 82
* DNV does not provide overall return periods for the trip extension sensitivity analysis and thus return periods were calculated
based on Brandsæter and Hoffman (2010).
Mitigation Measures and Mitigated Spill Return Periods
The next step in the DNV analysis estimates the impact of various mitigation measures
on spill return periods. DNV examines several risk reduction measures qualitatively
including enhanced navigational aids, vessel traffic management system, environmental
limits for safe operation, and the establishment of places of refuge along tanker routes.
The marine shipping QRA only quantitatively evaluates a single mitigation measure: the
tug escort plan.
DNV obtains the risk reducing effect of escort tugs from a confidential study it
completed in 2002 and applies the downward adjustments directly to powered
grounding, drift grounding, and collision incidents for the ENGP (Table A-­‐13). DNV
provides no details on the confidential study in the ENGP regulatory application.
93

Table A-­‐13: Estimated Reduction in the Frequency of Incidents from Tug Use
Incident type Condition Reduction Effect
Powered
Grounding
Laden with close and tethered tug
Laden with close escort
Ballast with close escort
80%
Drift
Grounding
Laden with close and tethered tug
Laden with close escort
Ballast with close escort
90%
80%
Collision Laden or ballast with close and/or tethered escort 5%
Source: Brandsæter and Hoffman (2010 p. 8-­‐120).
Table A-­‐14 compares mitigated return periods to unmitigated return periods for oil or
condensate spills. For all tanker spills, mitigation measures increase the spill return
period from 78 years to 250 years. Mitigation measures for powered and drift
grounding increase return periods by three-­‐ and four-­‐fold, respectively. DNV does not
identify mitigation measures for tanker foundering incidents and incidents involving
fire and/or explosion.
Table A-­‐14: Comparison of Unmitigated and Mitigated Oil/Condensate Spill Return Periods
Incident Type
Unmitigated
Return Period
(in years)
Mitigated
Return Period
(in years)
Powered Grounding* 107 480
Drift Grounding* 535 2,758
Collision* 1,084 1,138
Foundering* 25,821 25,821
Fire/ Explosion* 1,838 1,838
Total 78 250
Source: Brandsæter and Hoffman (2010).
* Calculated based on Brandsæter and Hoffman (2010).
Note: DNV does not identify mitigation measures for foundering and fire/explosion spills.
DNV also estimates the impact of mitigation measures on return periods by size of the
spill. As shown in Table A-­‐15, the use of tugs increases return periods from 200 to 550
years for an oil spill greater than 5,000 m 3 .
Table A-­‐15: Comparison of Unmitigated and Mitigated Return Periods for Various Oil Spill Sizes
Spill Volume (m 3)
Unmitigated
Return Period
(in years)
Mitigated
Return Period
(in years)
> 5,000 200 550
> 20,000 1,750 2,800
> 40,000 12,000 15,000
Source: Brandsæter and Hoffman (2010 p. 7-­‐110; 137).
2. Spills from Operations at Kitimat Terminal
DNV determines spill return periods for particular types of incidents at Kitimat
Terminal in the Marine Shipping Quantitative Risk Analysis. Furthermore, the Bercha
94

Group estimates risks from spills to the public and workers at the terminal in the
Northern Gateway Kitimat Terminal Quantitative Risk Analysis.
Spill Return Periods for Incidents at Kitimat Terminal
DNV prepares a risk analysis examining the likelihood of spills at the marine terminal.
Spill return periods for berthing/deberthing and cargo transfer operations at the
marine terminal largely follow the methodology for tanker spills. The approach
adopted by DNV estimates unmitigated and mitigated spill return periods at the
marine terminal by:
• Determining base terminal incident frequencies
• Determining conditional spill probabilities from an assessment of consequences
• Estimating unmitigated spill return periods
• Identifying mitigation measures and applying measures in the calculation of
mitigated spill return periods.
Base Incident Frequencies
DNV determines incident frequencies for berthing/deberthing and cargo transfer
operations at the marine terminal based on LRFP data and DNV’s research of operating
terminals in Norway. According to DNV, “Incident frequencies from terminals in
Norway are most representative of the operation planned for Kitimat Terminal and
should provide an appropriate forecast of the possible incident frequency at the
Kitimat Terminal” (Brandsæter and Hoffman 2010 p. 5-­‐71). DNV provides no evidence
to support the comparison between terminals in Norway and Kitimat Terminal.
DNV examines four types of potential incidents at the marine terminal:
1. A tanker is impacted by a harbour tug,
2. A tanker strikes the pier during berthing,
3. A tanker is impacted by a passing vessel, and
4. Various loading and/or discharging failures during cargo transfer operations.
According to DNV, the scenario whereby a tanker is impacted by a harbor tug is not
applicable to the ENGP. Given the small mass of a 600-­‐tonne tug boat, DNV claims that
it will not achieve the speed of over 3.9 metres per second (7.6 knots) required to
generate the 5 megajoules of energy required penetrate the hull of a tanker
(Brandsæter and Hoffman 2010 p. 5-­‐71). Therefore, DNV does not consider a tugboat
colliding with a tanker in the marine shipping QRA.
In the event that a tanker strikes the pier during berthing, DNV assigns a base incident
frequency of 6.3E-­‐05 derived from LRFP data and world tanker operations. A tanker
could strike the pier during berthing if tugs experience a mechanical failure or if human
error results in a loss of control of the tanker or tug. According to DNV, this scenario is
only applicable to tankers arriving at Kitimat Terminal and thus possible spills from a
tanker striking the pier will be condensate and/or bunker fuel from laden condensate
tankers or bunker fuel from oil tankers arriving to the terminal in ballast (Brandsæter
and Hoffman 2010 p. 6-­‐90).
95

For the third terminal incident type, DNV calculates the probability of a tanker being
struck at berth by a passing vessel with downward adjustments to a base incident
frequency. DNV estimates the frequency for a tanker being struck while at berth where
B is a base frequency of 9.0E-­‐06 obtained from DNV (2006), r1 is a 10% reduction factor
since Kitimat Terminal is not located in a high traffic port, r2 is a second reduction factor
of 50% to represent the effect of limited traffic forecast to pass Kitimat Terminal in the
open Kitimat Harbour combined with tug boats used for berthing 33 , t is the mean time
of 30 hours for loading/discharging operations at Kitimat Terminal, and v is the number
of vessels passing the terminal each year:
F striking = B *r 1 *r 2 *[t / (365* 24h)]* v
To adjust for the expected total number of tankers loading/discharging at Kitimat
Terminal per year, DNV multiplies Fstriking by 220 to represent an expected frequency of
6.8E-­‐05 per year for tanker traffic accessing Kitimat Terminal.
Finally, DNV estimates the probability of a spill from loading and discharge operations
associated with cargo transfer operations based on its Safety Analysis Handbook (DNV
2000 as cited in Brandsæter and Hoffman 2010). Spill probabilities for cargo transfer
operations represent several operational failures including overfilling of cargo tanks,
damage to loading arms or piping from external impacts such as mooring failure and
operator error, and leaks from loading arms or piping from internal impacts such as
corrosion or fatigue (Brandsæter and Hoffman 2010 p. 5-­‐74).
Conditional Probabilities
DNV determines the consequences of an incident during berthing/deberthing and cargo
transfer operations. Similar to conditional probabilities estimated for tanker spills,
conditional probabilities for terminal spills are the probability of a spill given that a
terminal incident has occurred.
Table A-­‐16 presents probabilities of a spill given that a tanker struck a pier during
berthing or a passing vessel impacts a tanker. Consequences from a tanker striking the
pier depend on where damage is inflicted on the tanker and if the damage results in
penetration of the cargo or fuel tank. DNV suggests that impact from the marine
structure is likely to result in penetrations to the hull above water and that, since it is
assumed that small indents do not penetrate the outer hull, the conditional probability
of a spill resulting in minor damage does not exist for laden tankers and tankers in
ballast (Brandsæter and Hoffman 2010 p. 6-­‐90). DNV estimates that the total
conditional probability of a spill given a laden tanker and tanker in ballast strike the
pier is 1% and 0.25%, respectively (Brandsæter and Hoffman 2010 p. 7-­‐112).
33 DNV references a confidential document it prepared in 2006 as the source of this reduction. The
confidential document is not appended to the marine shipping QRA and DNV provides no further information
about the reduction factor.
96

In the event a tanker is struck by a passing vessel, vessels will be travelling at low
speeds and although a tanker moored at Kitimat Terminal could be exposed to a side
impact, a tanker is not likely to experience a perfectly perpendicular side impact
(Brandsæter and Hoffman 2010 p. 6-­‐91). DNV assumes that tankers moored at the
marine terminal will be half full and thus the conditional probability of a spill is the
average for laden vessels and vessels in ballast presented in Table A-­‐16 for the tanker
collision scenario (Brandsæter and Hoffman 2010). DNV predicts some release of cargo
11% of the time there is an impact by a passing vessel.
Table A-­‐16: Damage Frequencies and Conditional Probabilities for Marine Terminal Incidents
Incident Damage LRFP Frequency Conditional Probability of Spill (%)
Type Category Distribution (%)
Laden Ballast
Minor 99 0 0
Tanker Major 1 100 25
Striking Pier Total Loss Negligible -­‐ -­‐
Total Conditional Probability 1 0.25
Impact by
Passing
Vessel
Minor 74.5
Major 25.5
Total Loss Negligible
Total Conditional Probability
0
42.5
100
11
Source: Brandsæter and Hoffman (2010 p. 6-­‐90 – 6-­‐91).
Probabilities and distribution of cargo release per loading/discharging operation are
summarized in Table A-­‐17. DNV develops a spill size distribution for each event type
from a review of historical databases from Norway and the UK and categorizes spill
sizes according to the International Tanker Owners Pollution Federation Limited 34
(Brandsæter and Hoffman 2010 p. 6-­‐91). DNV does not estimate large spills over 1,000
m 3 from an incident associated with cargo transfer operations.
Table A-­‐17: Spill Probability and Spill Size Distribution for Cargo Transfer Operations
Event Type
Probability of
Spill
(per operation)
Distribution of Spill Size
Small
(< 10 m3) Medium
(10 – 1,000 m3) Release from Loading Arm 5.1E-­‐05 90% 10%
Equipment Failure 5.1E-­‐06 0% 100%
Failure in Vessels Piping System or Pumps 7.2E-­‐06 90% 10%
Human Failure 7.2E-­‐06 90% 10%
Mooring Failure 3.8E-­‐06 0% 100%
Overloading of Cargo Tank* 1.2E-­‐04 0% 100%
Source: Brandsæter and Hoffman (2010 p. 5-­‐75) and DNV (2000) as cited in Brandsæter and Hoffman (2010 p. 6-­‐92).
Note: Spill Probabilities represent loading and discharge for every event type with the exception of overloading of cargo tank.
* Overloading of cargo tank can only occur during loading operations.
The volume of oil spilled from cargo transfer operations depends on several factors.
DNV uses the following equation to determine spill volumes from the failure of a single
34 According to Brandsæter and Hoffman (2010 p. 6-­‐91), the International Tanker Owners Pollution
Federation Limited categorizes spills as follows: small spills (700 tonnes ~ 1,000m 3).
97

loading arm where V is the volume spilled, T is the transfer rate, dt is the detection time
and E is the emergency shutdown time (Brandsæter and Hoffman 2010):
V = T *(dt + E)
Based on the above equation, the failure of one loading arm operating at its full rate of
transfer will result in different spill volumes for condensate and crude oil due to
different transfer rates and emergency shutdown times. The failure of a single loading
arm will result in a condensate spill of 200 m 3 or a crude oil spill of 250 m 3
(Brandsæter and Hoffman 2010 p. 6-­‐92) 35 .
Unmitigated Spill Return Periods
DNV combines incident frequencies and conditional probabilities to determine the
unmitigated risk evaluation of incidents at the marine terminal. Unmitigated risk is
presented as return periods for spills and is calculated based on forecast tanker traffic
calls to Kitimat Terminal and, where applicable, spill size distributions for particular
incident types.
In the event that a tanker strikes the pier during berthing, DNV claims that energy
produced from impacts to the pier is not sufficient to penetrate the outer or inner hull
of a tanker and thus results in a low likelihood that a spill will result from a striking
incident (Brandsæter and Hoffman 2010 p. 7-­‐112). DNV estimates that a crude tanker
in ballast or laden condensate tanker will strike the pier at Kitimat Terminal once in 72
years, whereas a tanker striking the pier and spilling cargo will occur only once in
14,600 years (Table A-­‐18).
Table A-­‐18: Frequency and Return Periods for a Tanker Striking the Pier During Berthing
Cargo Type
Number of
Approaches
Strikings
per Year
Striking Return
Period**
(in years)
Spill Frequency
per Year
Spill Return
Period
(in years)
Crude* 149 9.4E-­‐03 107 2.4E-­‐05 42,600
Condensate 71 4.5E-­‐03 224 4.5E-­‐05 22,300
Total 220 1.4E-­‐02 72 6.8E-­‐05 14,600
Source: Brandsæter and Hoffman (2010 p. 7-­‐112).
Figures might not add due to rounding.
* A release of crude includes the possible release of bunker fuel.
** Return periods for striking incidents are calculated based on the number of strikings per year.
With regards to a tanker impact by a passing vessel, DNV claims that the unmitigated
risk is “very low” (Brandsæter and Hoffman 2010 p. 7-­‐112) due to the low volume of
vessel traffic passing Kitimat Terminal. To support this claim, DNV multiplies the
annual frequency of 6.8E-­‐05 per year by the conditional probability of 11% to obtain
the annual probability of 7.3E-­‐06 for a tanker impact by a passing vessel that results in
35 DNV bases spill volumes of 200 m 3 of condensate and 250 m 3 of crude oil on a transfer rate of 3,000 m 3 per
hour for condensate and 4,000 m 3 per hour for crude oil, a 180 second detection time for both products, and
an emergency shutdown time of 60 seconds for condensate and 40 seconds for crude oil (Brandsæter and
Hoffman 2010 p. 6-­‐92).
98

a spill. Expressed as a return period, a spill from a tanker being struck by a passing
vessel will occur once every 130,000 years (Brandsæter and Hoffman 2010 p. 7-­‐112).
DNV bases unmitigated return periods for oil/condensate spills on the
loading/discharging of 220 tankers 36 and estimates unmitigated return periods for oil
spills according to the 149 oil tankers to call at Kitimat Terminal. As shown in Table A-­‐19,
DNV estimates that unmitigated small or medium size oil/condensate spills from cargo
transfer operations will occur once every 29 years, whereas a small or medium size oil
spill will occur every 34 years (Brandsæter and Hoffman 2010 p. 7-­‐113; 136).
Table A-­‐19: Unmitigated Return Periods for Oil/Condensate and Oil Spills from Cargo Transfer
Operations
Cargo Transfer Operation
Unmitigated Return Period (in years)
Small Medium Overall
Release from Loading Arm
Oil/Condensate
Oil Only*
99
146
890
1,316
89
132
Equipment Failure
Oil/Condensate
Oil Only*
-­‐
-­‐
890
1,316
890
1,316
Failure in Vessels Piping Oil/Condensate 700 6,300 630
System or Pumps
Oil Only* 1,036 9,321 932
Human Failure
Oil/Condensate
Oil Only*
700
1,036
6,300
9,321
630
932
Mooring Failure
Oil/Condensate
Oil Only*
-­‐
-­‐
1,100
1,766
1,100
1,766
Overloading of Cargo Tank
Oil/Condensate
Oil Only*
-­‐
-­‐
56
56
56
56
Total
Oil/Condensate
Oil Only
77
110
46
49
29
34
Source: Brandsæter and Hoffman (2010 p. 7-­‐113; 136).
* Return periods for oil spills are calculated with information from Brandsæter and Hoffman (2010).
Note: Small spills represent spills < 10 m3 and medium spills represent spills from 10 to 1,000 m3. Mitigation Measures and Mitigated Spill Return Periods
DNV applies mitigation measures to its unmitigated risk evaluation and recalculates
return periods for spills at the marine terminal. DNV quantitatively evaluates a closed
loading system with vapour recovery as the principal mitigation measure for spills at
the marine terminal. According to DNV, a closed loading system and vapour recovery
unit will “virtually eliminate tank overfilling” (Brandsæter and Hoffman 2010 p. 131)
thus reducing a spill to a negligible risk.
Table A-­‐20 presents mitigated return periods for oil/condensate and oil spills for cargo
transfer operations. According to DNV, a closed loading system and vapour recovery
unit installed at Kitimat Terminal can redirect excess oil from overfilled cargo tanks
into alternate ship tanks (Brandsæter and Hoffman 2010 p. 131). In the event that oil is
spilled by overloading cargo tanks and not recovered by the vapour return system, the
36 According to DNV, The only exception is the event in which cargo tanks are overloaded since the tanks can
only be overfilled during loading operations.
99

deck containment system would capture and direct the excess oil to vessel slop tanks
(Brandsæter and Hoffman 2010 p. 131). DNV claims that the closed loading system and
vapour recovery unit will eliminate any risk of an oil spill associated with overloading
cargo tanks and thus the probability per year of this event occurring drops to zero
(Brandsæter and Hoffman 2010 p. 131).
Table A-­‐20: Mitigated Return Periods for Oil/Condensate and Oil Spills from Cargo Transfer
Operations
Cargo Transfer Operation
Mitigated Return Period (in years)
Small Medium Overall
Release from Loading Oil/Condensate 99 891 89
Arm
Oil Only* 146 1,316 132
Equipment Failure
Oil/Condensate
Oil Only*
-­‐
-­‐
891
1,316
891
1,316
Failure in Vessels Piping Oil/Condensate 701 6,313 631
System or Pumps Oil Only* 1,036 9,321 932
Human Failure
Oil/Condensate
Oil Only*
701
1,036
6,313
9,321
631
932
Mooring Failure
Oil/Condensate
Oil Only*
-­‐
-­‐
1,196
1,766
1,196
1,766
Overloading of Cargo Oil/Condensate -­‐ -­‐ -­‐
Tank
Oil Only* -­‐ -­‐ -­‐
Total
Oil/Condensate
Oil Only
77
110
294
430
62
90
Source: Brandsæter and Hoffman (2010 p. 132; 136).
* Return periods for oil spills are calculated with information from Brandsæter and Hoffman (2010).
Note: Small spills represent spills < 10 m3 and medium spills represent spills from 10 to 1,000 m3. Table A-­‐21 compares the effects of mitigation measures for cargo transfer operations to
overall unmitigated spill return periods. According to DNV, mitigation measures more
than double the return period for unmitigated oil/condensate spills from 29 to 62
years.
Table A-­‐21: Comparison of Overall Unmitigated and Mitigated Oil/Condensate Spill Return
Periods for Cargo Transfer Operations
Cargo Transfer Operation
Unmitigated Return
Period (in years)
Mitigated Return
Period (in years)
Release from Loading Arm 89 89
Equipment Failure 891 891
Failure in Vessels Piping System or Pumps 631 631
Human Failure 631 631
Mooring Failure 1,196 1,196
Overloading of Cargo Tank* 56 n/a
Total 29 62
Source: Brandsæter and Hoffman (2010).
Rupture Failure Frequencies for Components at Kitimat Terminal
The Bercha Group (2012a) prepared a risk analysis report examining spills associated
with operations at the marine terminal. The Bercha Group submitted the report in May
100

2012 as a response to an information request from the Joint Review Panel for the ENGP
(MacDonald 2012). The risk assessment examines worker and public safety at the
marine terminal measured according to the risk to individuals exposed to oil or
condensate releases (Bercha Group 2012a). The report does not examine risks to the
environment or property surrounding the terminal.
The risk analysis uses a methodological approach that includes identifying hazard
scenarios, determining failure frequencies that result in a spill, evaluating the
consequences of spills, assessing risks, and recommending measures to mitigate risk
(Bercha Group 2012a). The risk analysis examines failure frequencies that include the
following terminal components: piping, pumps, storage tanks, and loading/unloading
arms (Bercha Group 2012a). Based on calculations from the Bercha Group report,
return period for a rupture from one of the terminal components is 282 years (Table A-­‐
22). The volume of oil or condensate spilled from a leak or full rupture at one of the
terminal components ranges between 10 m 3 and 80,000 m 3 depending on the type of
terminal component affected, the number of terminal components involved, and other
assumptions used by the Bercha group to model failure frequencies at the marine
terminal (Bercha Group 2012a).
Table A-­‐22: Rupture Failure Frequencies for Marine Terminal Components
Component
Rupture Failure
Frequency
Estimated Return Period
(in years)*
Piping 1.00E-­‐07 10,000,000
Pumps 1.00E-­‐04 10,000
Oil Tanks 5.50E-­‐05 18,182
Condensate Tanks 1.50E-­‐05 66,667
Loading Arm -­‐ Oil 2.29E-­‐03 437
Unloading Arm -­‐ Condensate 1.09E-­‐03 917
Total*
Source: Bercha Group (2012a p. 4.4).
3.55E-­‐03 282
* We calculate return periods based on information and data from Bercha Group (2012a).
The Bercha Group (2012a) concludes that risks to individuals from operations at the
marine terminal are within tolerable limits. Based on Individual Specific Risks
thresholds, the authors determine that risks to the public in the area surrounding the
terminal are insignificant at less than one in one million, while risks to workers are
higher although tolerable at a range of between one in 10,000 and one in 100,000. The
authors recommend mitigation measures to ensure safe operational and work
procedures at the marine terminal.
3. Spills from Onshore Pipeline Operations
The ENGP regulatory application contains several pipeline spill estimates. In its
regulatory application submitted in 2010, Enbridge prepared a section in Volume 7B
estimating return periods for spills that occur along the pipeline route. In 2012,
WorleyParsons and the Bercha Group submitted additional risk analyses estimating
pipeline spill likelihood for the project.
101

Spill Return Periods for Medium and Large Spills
Volume 7B of the Enbridge application identifies return periods for a hydrocarbon 37
release that occurs along the pipeline route. Enbridge calculates spill return periods
with liquid pipeline failure frequency data from 1991 to 2009 from the National Energy
Board (NEB). According to Enbridge, NEB data best represent the ENGP because data
are based on hydrocarbon transmission lines under the jurisdiction of the NEB
(Enbridge 2010b Vol. 7B, p. 3-­‐1). Enbridge notes that NEB data include pipelines up to
50 years old that use older technology and building material standards associated with
an increased frequency of failures compared to modern pipelines (Enbridge 2010b Vol.
7B, p. 3-­‐1). Accordingly, Enbridge factored improved design features and mitigation
planning characteristic of modern pipelines into its failure frequency calculations.
Improvement factors are expected to decrease the probability of pipeline failure
relative to NEB failure frequency data (Enbridge 2010b Vol. 7B, p. 3-­‐1). The ENGP
pipeline spill analysis does not provide the original or adjusted NEB spill data that was
used in the analysis, does not document the adjustments made to the NEB data and
provides no evidence to support the downward adjustment of NEB spill rates for the
ENGP.
Enbridge estimates return periods for various spill size categories of hydrocarbon
releases for six regions along the pipeline route. Enbridge bases spill size categories on
NEB classifications whereby a small spill is less than 30 m 3 , a medium spill ranges
between 30 and 1,000 m 3 , and a large spill exceeds 1,000 m 3 (Enbridge 2010b Vol. 7B).
Enbridge identifies the six physiographic regions along the pipeline route as the Eastern
Alberta Plains, the Southern Alberta Uplands, Alberta Plateau, Rocky Mountains,
Interior Plateau, and the Coast Mountains.
Table A-­‐23 presents spill return periods calculated by Enbridge for medium and large
spills in the six regions. Based on the adjusted NEB frequency data, spill return periods
for medium spills range between 118 and 1,082 years for each physiographic region,
while spill return periods for large spills range between 275 and 2,525 years per region.
Enbridge does not combine spill return periods for the six regions nor does it combine
return periods for medium and large spills. Combined, the six regions result in spill
return periods for medium and large spills of 41 and 95 years, respectively, for the
entire pipeline route. Similarly, combining overall return periods for medium and large
spills for the entire length of the pipeline results in a return period of 28 years.
Enbridge neither estimates nor presents spill return periods for small spills, yet simply
states “Smaller releases (less than 30 m 3 ) occur more frequently” (Enbridge 2010b Vol.
7B, p. 3-­‐2).
37 Enbridge (2010b Vol. 7B p. 3-­‐1) refers to pipeline spills as hydrocarbon spills. We interpret hydrocarbon
spills to include both oil and condensate and thus assume spill return periods represent oil or condensate
spills.
102

Table A-­‐23: Return Periods for Spills from the Northern Gateway Pipeline
Physiographic Region
Approximate Pipeline
Length (km)
Spill Return Period (in years)
Medium Large
Eastern Alberta Plains 166 287 669
Southern Alberta Plains 350 136 317
Alberta Plateau 44 1,082 2,525
Rocky Mountains 103 462 1,079
Interior Plateau 404 118 275
Coast Mountains 105 454 1,058
Total* 1,172 41 95
Combined Medium and Large Spills**
Source: Based on Enbridge (2010b Vol. 7B p. 3-­‐2).
28
* We combine spill return periods for each segment to represent the spill return period for the entire pipeline route.
** We combine return periods for medium and large spills into a single return period based on information provided in Volume 7B
of the ENGP regulatory application.
Note: Medium spills represent spills from 30 to 1,000 m3 and large spills represent spills > 1,000 m3. Spill Return Periods for Pipeline Ruptures Resulting in an Uncontrolled Spill
In addition to Volume 7B, Enbridge contracted WorleyParsons to prepare a risk
assessment evaluating the likelihood of a pipeline rupture that releases dilbit in an
unconstrained manner (referred to as full-­‐bore spill) (WorleyParsons 2012). The
report entitled Semi-­‐Quantitative Risk Assessment uses the following approach to
evaluating risk:
• Identify hazards that threaten the integrity of the pipeline system
• Estimate failure frequencies based on failure frequency modeling and expert
judgment
• Determine areas along the pipeline that have higher consequences from a full-­‐
bore rupture
• Examine the severity of unmitigated risk based on the frequency and
consequence of the rupture (WorleyParsons 2012).
The WorleyParsons (2012) report determines spill frequencies for leaks and full-­‐bore
pipeline ruptures over the length of the ENGP pipeline 38 . The analysis uses several
methodological approaches and datasets including a failure frequency model that uses
data from recently constructed pipelines, particularly Enbridge Line 4, as well as
pipeline spill data from the US Department of Transportation Pipeline and Hazardous
Materials Safety Administration (PHMSA) from 2002 to 2009 (WorleyParsons 2012).
The report evaluates eight hazards related to pipeline ruptures including internal
corrosion, external corrosion, materials and manufacturing defects, construction
defects, equipment failure, incorrect operations, damage from third parties, and
geotechnical and hydrological threats (Table A-­‐24). The WorleyParsons report
determines that the probability of a 594-­‐bbl (94 m 3 ) oil pipeline leak is 0.249, which
results in a return period of 4 years. For full-­‐bore ruptures releasing oil in an
unconstrained manner, the WorleyParsons report estimates an annual probability of
0.0042 (return period of 239 years) for a pipeline rupture releasing 14,099 bbl (2,242
38 We acknowledge that the pipeline distance of 1,176 km used in the WorleyParsons report differs from the
1,172 km used in Volume 7B of the ENGP regulatory application. Neither report addresses the discrepancy.
103

m 3 ) of oil. In addition to estimating oil spills, WorleyParsons estimates return periods
for a 593 bbl (94 m 3 ) condensate leak of 4 years and a 5,183 bbl (823 m 3 ) condensate
rupture of 273 years.
Table A-­‐24: Return Periods for Leaks and Full-­‐bore Ruptures
Return Period (in years)
Pipeline Hazard
Oil Pipeline Condensate Pipeline
Leak Rupture Leak Rupture
Internal corrosion 0 0 0 0
External corrosion 0 0 0 0
Materials and manufacturing defects 652 652 652 652
Construction defects 39 0 39 0
Equipment failure 5 0 5 0
Incorrect operations 47 0 47 0
Damage from third parties 436 1,308 1,385 4,149
Geotechnical and hydrological threats 0 530 0 530
Total 4 239 4 273
Source: WorleyParsons (2012); WorleyParsons (2012 as cited in WrightMansell 2012)
WorleyParsons also identifies areas along the pipeline corridor that are likely to suffer
higher consequences in the event of a full-­‐bore pipeline rupture (Table A-­‐25). According
WorleyParsons (2012), the six areas of high consequence represent a total of 181 km of
the 1,176 km ENGP, or approximately 15%, and a rupture is forecast to occur in a high
consequence area once every 1,260 years. Based on return periods, Kitimat River is
the most likely area affected by an unconstrained rupture due to geohazards in the
region. According to WorleyParsons (2012 p. 34), geohazards represent the most
significant threat to the Northern Gateway pipeline system.
Table A-­‐25: Return Periods for a Full-­‐bore Pipeline Rupture in High Consequence Areas
High Consequence Area
Length of Pipeline
Affected (in km)
Return Period
(in years)
Kitimat River 29 2,200
Athabasca River 32 12,000
Smoky River 33 12,000
Missinka River 41 14,000
Morice River 24 16,000
Gosnell River 22 24,000
Total
Source: WorleyParsons (2012 p. 29).
181 1,260*
* We estimate the total return period for all six high consequence areas.
Failure Frequencies for Pump Stations and Ruptures Along in Densely Populated Areas
The Bercha Group prepared two risk assessments examining separate risks from pump
stations and pipeline ruptures associated with the ENGP pipeline. Both studies use a
similar methodological approach that includes determining hazards, assessing the
frequency of occurrence of a spill, evaluating spill consequences, assessing risk, and
identifying mitigation measures that reduce risk (Bercha Group 2012b; 1012c).
104

The first Bercha Group (2012c) study assesses worker and public safety risks from an
oil or condensate spill at Smoky River Station, the largest pump station along the
pipeline route in Alberta near Grande Prairie. Table A-­‐26 shows the failure frequencies
per year for combined piping and pump spills at the Smoky River Station. The Bercha
Group obtained piping and pump failure frequencies from a report entitled Guidelines
for Quantitative Risk Assessment produced by Ministry of Housing, Spatial Planning and
the Environment in the Netherlands and uses incident statistics from the most recent
decade of data (Bercha Group 2012c p. 4.1). Based on failure frequencies, the return
periods for oil or condensate leaks or ruptures range between 332 and 4,082 years. To
estimate consequences of spills at Smoky River station, the Bercha Group (2012c)
estimates the amount of oil and condensate leaked from a spill at 50 m 3 and the
amount of oil or condensate released as a result of a rupture at 300 m 3 or 1,000 m 3 ,
respectively. From its frequency and consequence assessment, the Bercha Group
(2012c) concludes that individual risks from oil or condensate leaks and ruptures are
within allowable limits.
Table A-­‐26: Frequencies for Piping and Pump Spills at Smoky River Station
Type of Spill
Failure Frequency
(per year)
Return Period
(in years)*
Release Volume
(m3) Oil Leak 3.01E-­‐03 332 50
Oil Rupture 6.02E-­‐04 1,661 1,000
Condensate Leak 1.22E-­‐03 820 50
Condensate Rupture
Source: Bercha Group (2012c).
2.45E-­‐04 4,082 300
* Calculated from Bercha Group (2012c).
The second Bercha Group (2012b) study identifies areas along the pipeline route that
are known to have high public concentrations and determines the risk to individuals of
a pipeline spill at those locations. The authors identify three heavily populated areas
that could be affected by a pipeline spill: A casino near Whitecourt, AB; Burns Lake, BC;
and Kitimat, BC.
To estimate public risk from a pipeline spill, the Bercha Group (2012b) estimates the
failure frequency of the ENGP pipeline based on historical pipeline rupture data from
the NEB from 1991 to 2009 and makes several adjustments to the data to incorporate
various improvements in pipeline operations. As shown in Table A-­‐27, the failure rate of
0.372 per 10,000 km per year (10,000 km-­‐year) based on NEB historical data
decreases significantly to 0.109 spills per 10,000 km-­‐year due to numerous downward
adjustments ranging from 60 to 90%. The Bercha Group (2012b) concludes that risks
to public safety from a pipeline failure near the Casino at Whitecourt, Burns Lake, and
Kitimat are insignificant. Risks to residents and indoor and outdoor workers from a
pipeline releasing oil, condensate, or both are all below one in one million per year,
which the Bercha Group determines are acceptable levels of risk (Bercha Group
2012b).
105

Table A-­‐27: NEB Pipeline Failure Frequencies Adjusted Downward to Represent the ENGP
Pipeline
Failure
Causes
NEB Failure
Rate
(per 10,000
km-­‐year)
ENGP
Failure Rate
(per 10,000
km-­‐year)
Metal Loss 0.085 0.017 80%
Cracking 0.128 0.013 90%
External
Interference
Geotechnical
Failure
Material
Failure
0.021 0.021 -­‐ n/a
0.021 0.021 -­‐ n/a
0.021 0.009 60%
Other Causes 0.096 0.028 71%
Total 0.372 0.109 71%
Source: Bercha Group (2012b).
Downward
Adjustment Reason for Adjustment
Modern piggable pipelines are less
likely to fail than pipelines
represented in NEB data
Advanced coating technologies, low
stress design, and fatigue avoidance
through design and operations
Construction quality control,
inspection/testing after construction,
and operational surveillance
Estimated as same percentage as NEB
data (approx. 26%)
106